sensors

Review
Recent Advances in Bipedal Walking Robots: Review of Gait,
Drive, Sensors and Control Systems
Tadeusz Mikolajczyk 1, * , Emilia Mikołajewska 2,3 , Hayder F. N. Al-Shuka 4,5 , Tomasz Malinowski 1 ,
Adam Kłodowski 6 , Danil Yurievich Pimenov 7, * , Tomasz Paczkowski 1 , Fuwen Hu 8 , Khaled Giasin 9 ,
Dariusz Mikołajewski 10 and Marek Macko 11

1 Department of Production Engineering, Bydgoszcz University of Science and Technology,

Al. Prof. S. Kaliskiego 7, 85-796 Bydgoszcz, Poland; [email protected] (T.M.);
[email protected] (T.P.)
2 Department of Physiotherapy, LudwikRydygier Collegium Medicum in Bydgoszcz,
Nicolaus Copernicus University, 87-100 Torun, Poland; [email protected]
3 Neurocognitive Laboratory, Centre for Modern Interdisciplinary Technologies, Nicolaus Copernicus University,
87-100 Torun, Poland
4 Department of Aeronautical Engineering, Baghdad University, Baghdad 10001, Iraq;
[email protected]
5 School of Control Science and Engineering, Shandong University, Jinan 250100, China
6 Laboratory of machine Design, Lappeenranta University of Technology, 53850 Lappeenranta, Finland;
[email protected]
7 Department of Automated Mechanical Engineering, South Ural State University, Lenin Prosp. 76,
454080 Chelyabinsk, Russia
8 School of Mechanical and Material Engineering, North China University of Technology, Beijing 100144, China;
[email protected]
9 School of Mechanical and Design Engineering, University of Portsmouth, Portsmouth PO1 3DJ, UK;
Citation: Mikolajczyk, T.;
Mikołajewska, E.; Al-Shuka, H.F.N.; [email protected]
10 Institute of Computer Science, Kazimierz Wielki University, 85-064 Bydgoszcz, Poland; [email protected]
Malinowski, T.; Kłodowski, A.;
11 Faculty of Mechatronics, Kazimierz Wielki University, 85-064 Bydgoszcz, Poland; [email protected]
Pimenov, D.Y.; Paczkowski, T.; Hu, F.;
* Correspondence: [email protected] (T.M.); [email protected] (D.Y.P.)
Giasin, K.; Mikołajewski, D.; et al.
Recent Advances in Bipedal Walking
Abstract: Currently, there is an intensive development of bipedal walking robots. The most known
Robots: Review of Gait, Drive,
Sensors and Control Systems. Sensors
solutions are based on the use of the principles of human gait created in nature during evolution.
2022, 22, 4440. https://doi.org/ Modernbipedal robots are also based on the locomotion manners of birds. This review presents
10.3390/s22124440 the current state of the art of bipedal walking robots based on natural bipedal movements (human
and bird) as well as on innovative synthetic solutions. Firstly, an overview of the scientific analysis
Academic Editors: Abolfazl Zaraki
of human gait is provided as a basis for the design of bipedal robots. The full human gait cycle
and Hamed Rahimi Nohooji
that consists of two main phases is analysed and the attention is paid to the problem of balance
Received: 20 April 2022 and stability, especially in the single support phase when the bipedal movement is unstable. The
Accepted: 9 June 2022 influences of passive or active gait on energy demand are also discussed. Most studies are explored
Published: 12 June 2022 based on the zero moment. Furthermore, a review of the knowledge on the specific locomotor
Publisher’s Note: MDPI stays neutral characteristics of birds, whose kinematics are derived from dinosaurs and provide them with both
with regard to jurisdictional claims in walking and running abilities, is presented. Secondly, many types of bipedal robot solutions are
published maps and institutional affil- reviewed, which include nature-inspired robots (human-like and birdlike robots) and innovative
iations. robots using new heuristic, synthetic ideas for locomotion. Totally 45 robotic solutions are gathered
by thebibliographic search method. Atlas was mentioned as one of the most perfect human-like
robots, while the birdlike robot cases were Cassie and Digit. Innovative robots are presented, such
asslider robot without knees, robots with rotating feet (3 and 4 degrees of freedom), and the hybrid
Copyright: © 2022 by the authors.
robot Leo, which can walk on surfaces and fly. In particular, the paper describes in detail the robots’
Licensee MDPI, Basel, Switzerland.
propulsion systems (electric, hydraulic), the structure of the lower limb (serial, parallel, mixed
This article is an open access article
distributed under the terms and
mechanisms), the types and structures of control and sensor systems, and the energy efficiency of the
conditions of the Creative Commons robots. Terrain roughness recognition systems using different sensor systems based on light detection
Attribution (CC BY) license (https:// and ranging or multiple cameras are introduced. A comparison of performance, control and sensor
creativecommons.org/licenses/by/ systems, drive systems, and achievements of known human-like and birdlike robots is provided.
4.0/). Thirdly, for the first time, the review comments on the future of bipedal robots in relation to the

Sensors 2022, 22, 4440. https://doi.org/10.3390/s22124440 https://www.mdpi.com/journal/sensors

Sensors 2022, 22, 4440 2 of 31

concepts of conventional (natural bipedal) and synthetic unconventional gait. We critically assess
and compare prospective directions for further research that involve the development of navigation
systems, artificial intelligence, collaboration with humans, areas for the development of bipedal robot
applications in everyday life, therapy, and industry.

Keywords: robotics; bipedal locomotion; human gait; bird gait; synthetic-based biped gait; hu-
manoid; sensors

1. Introduction
Mobile robots of various locomotion mechanisms have revealed limited built-in auton-
omy that obtains information from both internal and external sensors for pre-planned and
purpose-oriented locomotion [1–3]. Legged robots offer greater possibilities than wheeled
and tracked robots in terms of working environments. Legged robots can move over
regular and irregular terrain without any hardware modifications and have demonstrated
exceptional mobility [4]. Generally, the two-legged robot mimics the way human moves.
It is intended to undertake a variety of tasks including civilian and military activities in
hazardous conditions, entertainment and education, and assistance for the elderly and the
disabled. Owing to the smaller foot contact area with the ground and the smaller number
of driving effector, bipedal walking robots’ total energy consumption may be lower in
comparison to the multi-legged robots [5]. A historical overview of leg-driven robots and
machines, as well as an introduction to the walking pattern and stability generators, are
provided in the articles by Bekey GA [6], Raibert MH [7], and Al-Shuka [8,9], respectively.
The information gathered by the robot sensors can be used for real-time analysis of the
environment. Moreover, there are cognitive robots such as the intelligent robot companions,
which may be helpful in the therapy of children with autism spectrum disorder (ASD),
depression, or as an aid for the elderly [10–13]. Generally, the common challenges of biped
robots include, but are not limited to, the following:
• Bipedal robots have unstable structures due to the passive joints located at the unilat-
eral contact between the foot and the ground [14–16];
• One-sided contact of the foot with the ground and a complex configuration of the gait
cycle bring about the highly non-lineartrajectory of the bipedal robot [14,17];
• Bipedal robots have multiple degrees of freedom (DOFs). Most researchers use simpli-
fied models to reach a trade-off between simplicity and the dexterity [18];
• Bipedal robots are most often designed to interact with unknown environments and
are expected to achieve a high level of autonomy [19,20];
• Simulation is required as a part of many control strategies for bipedal walking [19,21].
These topics are related to advanced mechanics, control theory, electronics, artificial
intelligence (AI), and the knowledge of human anatomy. Obviously, the research and
development of bipedal walking robots are truly interdisciplinary. To solve these problems,
the close cooperation of research teams from various fields is required. As far as we know,
the first walking robot is described in theIlliad, by Homer. A wooden walking device
named Mu Niu Liu Ma (in Chinese) was designed in the year 231. Modern studies on
humanoid robots began in the early 1960s with artificial hands and arms—for supporting
the physical work of men. In 1969, Vukabratovic et al. [22,23] developed several original
self-propelled exoskeletons to help paraplegics. However, the most well-known humanoid
robot is Asimo, created by Honda in the year 2000, based on prototype E0 (1986) [24].
Review articles on bipedal robots can be found in the scientific literature [25–30].
Wahde et al. [25] presented the research progress of biped and humanoid robotics in
the year 2002. The authors described both commercial and research projects showing
biologically inspired biped robots. Bezerra et al. [26] presented a review of the main types
of biped robots that developed until the year 2004. Silva et al. [4] provided a review in the
field of optimization methods for the construction and manners of movement of walking
Sensors 2022, 22, 4440 3 of 31

robots, which were characterized by greater energy consumption, as compared to robots on

wheels [27]. Ficht et al. [28] investigated an important aspect of the development of modern
walking robots—a humanoid robot with a fully 3D-printed structure supported by off-the-
shelf components. This method significantly reduces the implementation cost and allows
modifications of the structure. Ye et al. [29] presented a brief overview of the methods
thatenable stable walking and running in bipedal robots. Modern methods significantly
improve the endurance and adaptability of robots. They can traverse unknown terrain
with a variability of the ground exceeding 20% of the length of the legs. Bipedal robots can
regain balance after a sudden push, not only in a stationary state but also while in motion.
However, running is still a problem for robots.Ficht et al. [30] provided an overview on the
most advanced solutions for two-legged robots. The article presented the applied kinematic
structure of the leg drive of modern humanoid robots, and the directions of development
of bipedal robot control systems. Particular attention was paid to the technological aspects,
emphasizing the role of 3D printing in creating new solutions. Gupta and Kumar [31] have
made a good review paper concerning design models and control strategies for motion and
running of underactuated biped robots.There are interesting works on the analysis of the
dynamic walking of biped robots, for example, see the works of Westervelt et al. [32] and
Sadati et al. [33].
Regarding the broad topic of the bipedal walking robots, there is a twofold direction
of research, distinct from each other: (1) the progressive research about bipedal walking
robots, and (2) review studies (surveys of the literature) on the various research and to the
results obtained so far—a survey of the literature. In the present work, we accomplished
this in the second category. We are conscious that the current paper, despite more than
160 references cited, may address this paper more deeply—thus, it may be regarded as
introductory for further deeper and narrower studies in their focus. We hope it will inspire
researchers, engineers, and clinicians for further interdisciplinary studies paying attention
to their various detailed aspects of the very important issue of robotic bipedal walking,
e.g., low limbs exoskeletons, also controlled using a brain-computer interface as half body
neuroprosthesis.
The focus of this article is to survey the challenges that relate to the theory of human
and bird, design, drive, and control systems of bipedal robots. It considers the solutions
based on the natural biped motion (human-like and bird-like) and also synthetic, heuristic
solutions of bipedal walking machines. Section 2 presents the main biped gait issues of
humanoids and birds. Section 3 focuses on the implementation of self-developed bipedal
walking robots based onnatural gait patterns and syntheticsolutions. In Section 4, drive and
control methodology of biped walking robots, the sensing methods to support their two-
legged movement, and also energy efficiency are discussed. Section 5 shows the prospect
of bipedal walking robots and their potentials. Finally, the summary and conclusions are
presented in Section 6.

2. Bipedal Walking Mechanism

2.1. Bipedal Walking of Human
Walking requires integrated and coordinated activities of thenervous, muscular, and
skeletal systems, which control and enable the gait, as well as provide the closed-loop
feedback for reactive balancing and stability control. The main theories to explain the
bipedal walking are listed below [5,8,9]:
• The evolution theory;
• Theory of minimizing energy consumption;
• The theory of maturation (grow from childhood to adulthood);
• Central pattern generator theory;
• The theory of bipedal locomotion as a result of the two cooperating mechanisms;
• The theory of bipedal robots (gait can be generated by stability and online feedback
control [7].
Sensors 2022, 22, 4440 4 of 31

Each of the aforementioned theories has some supports, but none of them comprehen-
sively explains all possible physiological and pathological mechanisms. Researchers have
attempted to extract the bipedal robot design inspirations from their biological counterparts,
but there are still many problems regarding the gait complexity of different levels:
• The general view: the stable, controlled bipedal gait in various environments and
during fulfilling various meaningful tasks (including cognitive tasks);
• Combination and transition between thedifferent modes of bipedal locomotion: walk-
ing and running without falling [34];
• The high-level control of walking: cooperation of trajectory planning algorithms and
central pattern generators [35];
• The high-and low-level signal processing: translation of ground contact into signals
activating particular joints and muscles [36,37];
• The adaptive control layer that considers dynamic stability, detecting ground move-
ment or slippery surfaces during walking.
Many of the above-mentioned problems are possible to solve based on the theory
of nonlinear dynamical systems (stability/equilibrium analysis, cycle properties, multi-
fractal analysis) [35]. However, no one-size-fits-all criterion ensures the equilibrium of
bipedal robots.

2.2. Application of Bipedal Gait of Humans for Biped Robots

A better understanding of the bipedal walking robots comes from the analysis of
human walking behaviour adaptation strategies that are based on interlimb and intralimb
coordination. Goswami [38] analysed a biped robot’s postural stability and foot rotation.
Furthermore, analyses of passive walking with knees were presented in [39]. Early studies
of these mechanisms were based on a split-belt treadmill. A two-tier bipedal walking control
model was described by Fujiki et al. [40]. Its main principles were based on the following:
• Independent neural control of parameters during gait;
• The coupling of the metatarsal to the medullary regulation;
• The ability to adapt and remember the new inter-acting patterns;
• Involvement of the cerebellum in generating substantive commands based on foot signals;
• The interaction of the cerebellum with the brainstem is a key control for interlimb
coordination in slow walking [41–43].
Gait pattern is generalized to natural walking even if the vision is removed [44]. This has
been achieved thanks to evolution, experience, training, and trial-and-error practice [45].

2.3. Concepts of Bipedal Walking Robots Based on Human Walking

Humans and animals have matchless mobility abilities due to their versatile mechani-
cal reconfiguration capabilities. The features of animal and human locomotion systems are
analysed [46,47]. During the movement, humans use about 20 DOFs from all the 300 DOFs
of the musculoskeletal system [18,48,49]. Based on this observation, the following important
observations from previous literature are summarized:
• The majority of researchers try to adapt the idea of human walking using simplified
models [9]. Among possible choices there are 2-link model [8,50], 3-link model [51],
5-link model [17,52,53], 30-DOFs humanoid [54], etc.
• The complete gait cycle of human walking consists of two main successive phases:
the double support phase (DSP, 20%) and the single support phase (SSP, 80%) with
intermediate sub-phases (toe-off, forward swing, and heel strike) [54]. DSP results in a
closed chain mechanism, while the SSP starts when one of the feet begins the forward
swing phase [55].
• Balance and stability problem is important especially during the SSP when bipedal
motion is unstab [56]. In general, there are two kinds of stability criteria: static and
dynamic. Static stability depends on the vertical projection of the centre of mass on the
support surface [57] and allows to simplify the design of bipedal robots considerably
Sensors 2022, 22, 4440 5 of 31

and open-loop balancing can be assumed with large enough feet [58]. For dynamic
stability, the following methods are usually considered: zero moment point (ZMP) [57],
centroidal angular momentum [59], footstep-based criteria [60], and periodicity-based
gait [61–64]. Further details can be also found in previous literature [6,9,65].

Remark 1. Some researchers consider the ZMP-based locomotion as static motion as long as the
walking is slow in comparison with motion of periodicity-based stability. In effect, most passive
walkers lie within the category of periodicity-based motion. The passive walkers use minimum
actuators for motion on a plane surface or even without actuators but with an inclined surface for
walking. In contrast to active walkers that use full actuation or over actuation strategy for walking,
most of them are designed based on the ZMP, see [31,66] for more details.

Remark 2. The human walks using muscles and nerves as actuators and controller elements, respec-
tively. Without the muscles and nerves, the human behaves as a passive walker [64]. Consequently,
the bipedal robot can be modelled as an inverted pendulum with passive dynamic walking that
exploits dynamics only, e.g., McGeer’s passive walker [67,68]. In view of the above, bipedal robots
have been known to exhibit complex behaviour like chaos and bifurcations with hybrid algebraic-
differential equations [64,69]. Simple models were selected to investigate chaos and bifurcations
phenomenon such as compass-gait biped [70], point-feet straight legged robot [71], semi-passive
biped actuated in torso [72], an underactuated bipedal robot with constant torque being applied on
the support leg [73], a 3D passive walker [74], and rimless wheel models [75]. On the other hand,
thePoincaré map is a powerful tool to investigate the stability of passive dynamic walkers; however,
difficult computations arise in solution of complex biped systems. Therefore, Zengui et al. [69]
proposed time linearization of the hybrid bipedal system with a state feedback controller to stabilize
the linearized Poincaré map. In general, two basic methods are available for controlling chaos [76]:
the Ott–Grebogi–Yorke method and the delayed feedback control method, see [64] and the references
therein for more details.

One of the important issues of biped locomotion is the generation of the optimal
trajectory that provides the stability while avoiding collisions with obstacles. In practice,
several algorithms can be combined to generate bipedal locomotion patterns [6]:
• The learning process (requires intelligence);
• A considerable level of ability to adapt to different conditions or to solve tasks of
different obstacles in the terrain;
• Under certain conditions (e.g., long-distance walking), optimal movement to reduce
walking energy consumption.
Most designers and scientists proposed control systems for bipedal robots based
on predefined trajectories. Methods used for online ZMP compensation can be based
on preview control [77], model predictive control [65,78–82], or AI [83,84]. In contrast,
modern systems, based on artificial intelligence, can produce sound results without direct
modelling of the phenomena [85], although there are different approaches adopted to
generate balanced/stabilized walking patterns. Hayder A.S. et al. [9] present the classifi-
cation of gait generation approach based on model, biological mechanisms, and natural
dynamics [19,52,58,86].

2.4. Selected Issues of Bird Gait

In nature, several animals have been known to move perfectly on two legs such as
dinosaurs. Thulborn [87] gave the known relationships derived from mammals such as
speed, gait, and body size to determine the gaits and theoretical maximum speeds of
dinosaurs. He concluded that smaller bipedal dinosaurs ran at speeds of up to 35 or
40 km/h, “ostrich dinosaurs” 35–40 km/h, and up to 60 km/h. Larger bipedal dinosaurs
were walking or slowly trotting with speeds in the range of 15 to 20 km/h. According to
the theory of evolution, birds are one of the closest successors of the dinosaurs—they also
move on two legs independently of if they can fly or not. Alexander [88] analysed the gaits
Sensors 2022, 22, 4440 6 of 31

of different animals (reptiles, birds, and mammals). He concluded that mammals move
similarly when the Froude number is considered, as shown in Equation (1):

u2
F= (1)
gh

where: u is the speed, g denotes the acceleration, and h stands for the height of the hip.
In another paper [89], the author observed that, in addition to humans and birds
utilizing bipedal walking mechanisms, cockroaches can also run in bipedal mode. Many
birds walk and run with knees bent, back and femurs in a position near to horizontal. The
author showed the differences between the two-legged movement of humans and birds.
In humans, two peaks of force can be observed during walking, and a single peak when
running. Two peaks of force similar to that of human walking are also observed in birds.
Human walking is energy efficient, but human running is rather expensive. For birds,
bipedal locomotion is economical for wading birds, and expensive for geese and penguins.
Abourachid et al. [90] presented a short review of the biological bipeds to explain the
differences between the body structures of humans and birds. The authors proposed a bird
model scaled up to the same mass and height of the centre of mass as a humanoid model.
The results of such simulation showed the advantage of the bird model in comparison
with the humanoid model. Results have also confirmed the possibility of decomposition
of the bird’s system on the trunk and thighs as one part, and leg as a second part. The
authors noted that the movement of birds may be a good inspiration for building walking
robots. Daley et al. [91] observed that birds are ecologically diverse and span a large range
of body size and limb proportions, yet they all use their hind limbs for bipedal terrestrial
locomotion. The authors also indicated that birds use different strategies depending on
terrain such as:
• Independent control of the angular movement and the length of the legs to ensure
dynamic stability;
• Control of the speed of movement with positive feedback to ensure a constant load on
the legs in uneven terrain;
• Adjusting the muscles to the load, which stabilizes the mechanical energy usage of
the body;
• Complex regeneration strategies that allow changing the dynamics of the body, while
regulating the load on the legs, which in turn minimizes the risk of falling.
The authors [92] analysed the scaling of the gait of bipedal birds related to the load and
muscle power that results in adapting the frequency of the gait. The stability and agility of
movement were also analysed. At dynamically similar speeds, small birds use relatively
shorter stride lengths and higher stride frequencies as compared to large birds. Birds with
long legs as compared to their mass, use longer strides and lower swing frequencies. Birds
are the only living animals that can stand, walk, and run on two legs, just like a human. At
the same time, the fastest birds (ostriches) are much faster than humans. The movement
of birds is also very agile. This is due to the different structures of the kinematic system
and muscles in comparison to the human gait. Leveraging biological patterns of bird
movement is an inspiration for bipedal robot’s alternative [90] to humanoid robots that
have been developed for many years. The construction of bird-inspired bipedal robots will
be presented in the next section.

3. Designs of Biped Walking Robots

3.1. Overview of Bipedal Robots
To reveal the current state of the art, a search and review of the literature on bipedal
robots published up to 31 December 2021 was conducted using the following keywords:
“bipedal walking” and “bipedal robot”. The following electronic databases were searched:
Scopus, Pubmed/Medline, PeDro, Cochrane, Pro-Quest as far as MeRoDa (Medical Robotic
Database). After evaluation of publications according to their titles, abstracts, and full
Sensors 2022, 22, 4440 7 of 31

texts, a narrative review was performed. In a six-year window, all articles dealing with the
walking robots that can be found on Google Scholar represent around 10,000 titles. The
state of knowledge is based mainly on narrative reviews as well as on previous technical
reports. The design of bipedal walking robots presented in Table 1 can be divided into
solutions based on the following biological patterns:
• Human Biped Walking Robots (HBWR);
• Bird Biped Walking Robots (BBWR);
• Synthetic Biped Walking Robots (SBWR)—other solutions based on a heuristic, syn-
thetic ideas.

Table 1. Summary of previous bipedal walking robot projects.

No. Topic Years Type Remarks

Experimental biped robots
Artificial lower limb. The base for further studies on
1. WL-1 [93] 1966–1967 HBWR
bipedal robots.
Bipedal walking device Master/Slave: walking, sitting, and
2. WL-3 [93] 1968–1969 HBWR
standing device.
A bipedal walking robot with artificial rubber muscles,
3. WAP-1 [93] 1969 HBWR
pre-programmed gait sequence.
A bipedal walking robot with effectors, automated posture
4. WAP-2 [93] 1970 HBWR
adjustment thanks to feet sensors.
5. WAP-3 [93] 1971 HBWR First bipedal walking robot able to climb the stairs.
6. WL-5 [93] 1970–1972 HBWR Heavy bipedal walking robot with flexiblehips.
7. WL-9DR [93] 1980–1982 HBWR Quasi-dynamical robot. One step in every 10 s.
One step in every 4.4 s,
8. WL-10, 10R [93] 1982–1983 HBWR
possibility of turnaround.
9. WL-10RD [93] 1984 HBWR Dynamically stable robot. One step in every 1.3 s.
Walking robots based on natural bipeds
Interactive robot.
10. ASIMO [24] 1986 HBWR
Gait velocity up to 5.95 km/h.
11. HRP-2 [94] 2002 HBWR Lifting objects, moving in unknown terrain.
Open-source robotics humanoid robot for research of human
12. iCub [95] 2004 HBWR
cognition and artificial intelligence.
13. NAO [96] 2007 HBWR Small walking robot for educational tasks.
Female robot with a realistic face. Movement-based on captured
14. HRP-4C [97] 2009 HBWR
human motion.
15. HRP-4 [98] 2010 HBWR Can collaborate with humans, exhibits human-like gait.
16. PETMAN [99] 2011 HBWR Protection Ensemble Test Mannequin.
17. REEM-C [100] 2013 HBWR Human–robot interaction.
Search and rescue tasks, very dynamic, walking in uneven terrain,
18. ATLAS [101–103] 2013–2020 HBWR
running, jumping capabilities.
19. Robonaut 2 [104] 2014 HBWR NASA robots get special legs with manipulation functions.
TORO is a humanoid robot controlled by torque used to study
20. TORO [105] 2014 HBWR
bipedal walking and autonomous manipulation.
21. ATRIAS [106] 2015 BBWR Bipedal robot inspired by bird gait kinematics.
Rich sensory system control of loads and thermal sensing/fatigue
22. WALKMAN [107] 2015 HBWR
of actuators and electronics.
Sensors 2022, 22, 4440 8 of 31

Table 1. Cont.

No. Topic Years Type Remarks

23. CHIMP [108] 2015 HBWR Carnegie Melon University robot for rescue task.
THORMANG Open-source advanced walking robot with the possibility to
24. 2015 HBWR
[109,110] change to the wheeled platform.
25. Valkyrie [111] 2015 HBWR NASA’s Most Advanced Space Humanoid Robot.
This robot can use tools, open doors, drive a vehicle, and
26. DRC-Hubo+ [112] 2015 HBWR
transform into a wheeled robot.
27. DURUS [113,114] 2015 HBWR SRI’s robot with high energetic efficiency.
28. HBS-1 [115] 2016 HBWR Child size walking robots for different tasks.
Kenogro was equipped with body skeletal structure driven
29. Kenogro [116] 2016 HBWR
by muscle.
Hydra uses electro-hydrostatic actuators (EHAs) with its own
30. Hydra [117] 2016 HBWR
pump. It combines the advantages of hydraulic and electric drives.
31. Cassie [118] 2016 BBWR Dynamic robots walk and run as the animal (bird).
Adult-sized open-source, low cost, a 3D printable
32. NimbRo-OP2 [119] 2017 HBWR
humanoid robot.
TALOS is humanoid, which can walk on uneven terrain, and
33. TALOS [120] 2017 HBWR perform tasks both in research and industrial environments (can
operate power tools and lift 6 kg in each hand).
A humanoid robot that can use a power tool and manipulate
34. HRP-5P [121] 2018 HBWR
large objects.
Robots with many sensors based on Cassie kinematic for
35. Digit [122] 2019 BBWR dynamical running in difficult environments, can do
advanced tasks.
WANDERRER Walking robot with an innovative mechanism for high energy
36. 2020 HBWR
[123,124] performance and endurance.
Do-It-Yourself bipedal walking robots
DARwIn-OP, Dynamic Anthropomorphic Robot with Intelligence—Open
37. 2011 HBWR
DARwIn-OP2 [125] Platform Robot humanoid kit.
Low-cost 3D Printed
38. HBWR Cost lower than 1000 Euro.
Humanoid Robot
Robot humanoid kit Interactive robot
39. Poppy [126] 2012 HBWR
Open-source license.
40. Lim andYeap [127] 2012 HBWR 6 DOFs walking robot.
41. RQ-HUNO [128] 2014 HBWR Robot humanoid kit.
42. Red-Dragon V3 [129] 2014 HBWR Mobile device-controlled robot.
43. w00dBob [130] 2014 HBWR A biped wooden robot controlled by Arduino Nano.
Synthetic bipedal walking robots
44. RotoFoot * [131–134] 2014 SBWR Walking robot with rotary feet.
45. Slider [135,136] 2018 SBWR Walking robot without knees.
46. LEO [137,138] 2021 SBWR Multimodal walking robot with the possibility to fly as a drone.
* Name of robot proposed for use in this paper.

Humanoid robots are bipedal walking devices built to resemble human-like locomo-
tion. Do-It-Yourself bipeds are characterized by a simple kinematic chain design to reduce
actuation costs and simplify the control system; such designs are designed towards simple
manufacturing—often low-cost 3D printing or laser cutting of all the parts. Another typical
feature of such designs is their open-hardware and open-source licensing.
Sensors 2022, 22, 4440 9 of 31

3.2. Human Biped Walking Robots (HBWR)

Most of the solutions presented in Table 1 are designed by imitatingthe human walk-
ing mechanism. Two-legged robots mimicking human walking mechanisms are capable
of passing over or avoiding obstacles; however, they are characterized by complex con-
struction and control. For an efficient walking pattern, the bipedal robot needs 12 DOFs,
as in the best solutions of bipedal walking robots, for example ASIMO [24]. There are
attempts to simplify the kinematic structure of two-legged robots to 6 DOFs (Lim andYeap
walking robot [127]), which results in a moderate reduction of movement capabilities. The
robots presented up to this point are most often equipped with rotary or linear actuators
electrically driven by DC servos or stepper motors. Boston Dynamics, in 2013, unveiled
the prototype of the two-legged walking robot Atlas [101], powered by a hydraulic system.
Then in 2016 [102], the company presented an excellent design with a perfectly refined
hydraulic muscular system. This type of drive provides enormous strength and excellent
dynamics. Initially, the controls were in development and the robot’s gait was wobbly. In
2019 [103], Boston Dynamics presented Atlas with improved software, which does not
directly control the robot’s joints; instead, it uses pre-programmed relations between the
robot’s features and its relation to environment interaction. The Atlas robot uses the whole
body for balancing and performing acrobatic movements like jumping over obstacles or
back-flips [103]. Undoubtedly, the solution presented by the Boston Dynamics company
sets a completely new level in terms of the construction of walking robots.

3.3. Bird-Based Biped Walking Robots (BBWR)

Recently, new solutions for two-legged walking robots areput forward using biolog-
ical patterns that are not inspired by human gait, instead, they resemble birds walking
mechanics. The examples of such walking robots are Cassie [118] and Digit [122]. We can
see the difference between the structure of leg joints for human-based robots and Cassie
or Digit robots that are similar to birds’ lower bodies. The key feature of those robots is
excellent manoeuvrability and the possibility to move in difficult terrain. The Cassie robot
was a protoplast for the Digit, which, as an evolution design, was equipped with arms and
a head containing a LIDAR system for navigation [122]. The Cassie robot presents excellent
walking and jogging capabilities; in addition, it exhibits good manoeuvrability due to the
quick rotation [139]. These robots quickly reached a high level of design in comparison to
the best solution of humanoid robots due to the advances in control systems of humanoid
robots and using a unique drive design.

3.4. Synthetic Bipedal Walking Robots (SBWR)

The last section of Table 1 presented synthetic walking designs. The designs utilize
a cyclical leg shifting mechanism to maintain stability by ensuring the centre of gravity
projection on the floor stays underneath the grounded feet. Table 1 presents two such solu-
tions that combine the sliding motion to produce vertical leg movement when transferring
the load from one leg to the other. Wang et al. [135,136] proposed an innovative design
of a two-legged robot called SLIDER, which used a sliding joint in each leg to replace the
knee–hip rotary motion that was used in nature for leg lifting [135,136].
The design prepared at the Imperial College of London [135] uses linear movements of
the straight leg in the hip, replacing the effect of knee–hip combined rotation. The innovative
aspect of the SLIDER design is the compact and lightweight mechanism of vertical hip
movement. The proposed system of locomotion of the SLIDER robot uses only 5 DOFs per
leg—totally giving 10 DOFs. Simulated robot motion animations are presented in [136].
One of the challenges in two-legged robots is the execution of twists and turnaround
motions. The new innovative solution of walking mechanism—which is not based on hu-
man gait as most solutions of two-legged walking robots—was presented by
Mikołajczyk et al. [131–134]. This new walking mechanism uses two parallel legs sliding
in corps and is equipped with swivel feet as shown in Figure 1.
 DOFs per leg—totally giving 10 DOFs. Simulated robot motion animations are presented
in [136].
One of the challenges in two-legged robots is the execution of twists and turnaround
motions. The new innovative solution of walking mechanism—which is not based on hu-
man gait as most solutions of two-legged walking robots—was presented by Mikołajczyk
Sensors 2022, 22, 4440 10 of 31
et al. [131–134]. This new walking mechanism uses two parallel legs sliding in corps and
is equipped with swivel feet as shown in Figure 1.

(a) (b)
Figure 1. An
Figure 1. innovative
An innovativebipedal robot
bipedal with
robot swivel
with feet:
swivel DR,DDRL,—drives
feet: of feet,
DL —drives FR, FLR—swivel
of feet, feet feet
, FL —swivel of of
robo, DC—drive of robots leg, COG—centre of gravity, L—distance between feet axis, H—vertical
robo, DC —drive of robots leg, COG—centre of gravity, L—distance between feet axis, H—vertical
move of feet:
move (a) 3DOF
of feet: robot
(a) 3 DOF balancing
robot mass
balancing moved
mass using
moved DC drive;
using (b) 4DOF
DC drive; robotrobot
(b) 4 DOF balancing massmass
balancing
moving using DG—independent balancing drive.
moving using DG —independent balancing drive.

TheThe3DOF robot
3 DOF is equipped
robot is equipped withwiththreethree
drives (Figure
drives 1a): central
(Figure common
1a): central common drive–𝐃 𝐂
drive—
usesDgear control of moving legs in the vertical direction, drives of swivel
C uses gear control of moving legs in the vertical direction, drives of swivel feet control:
feet control:
right foot—𝐃
right foot—D 𝐑 and left foot—𝐃𝐋 , respectively,
R and left foot—D
on αR and αL angle. This way the robot is
L , respectively, on αR and αL angle. This way the robot
equipped with 3DOFs, in which the 1st DOF
is equipped with 3 DOFs, in which the 1st DOF is controlled is controlled by verticalbysimultaneous move-
vertical simultaneous
ments of the legs,
movements andlegs,
of the the remaining two DOFs
and the remaining twoare the rotations
DOFs of the feet
are the rotations parallel
of the to the to
feet parallel
ground surfacesurface
the ground (Figure(Figure
1a). For 1a).static stabilization,
For static gear gear
stabilization, connection to balancing
connection to balancingmass is is
mass
used to move
used to move thethe
centre of of
centre gravity—COG—from
gravity—COG—fromone onefoot
foottotoanother.
another. TheThe movement of the
the mass
mass in in the
the simplest
simplest version
version is is mechanically
mechanically combined
combined withwith the
the leg
leg motions
motions (Figure
(Figure 1a).
1a).The
Thesequence
sequenceofofaasingle singlerobot
robotstepstepisisasasfollows
follows(the
(therobot
robotstands
standsonontwotwofeet–starting
feet–starting from
fromthethe left
left foot):
foot):
• • TheThe startstart
of the
of central common
the central commondrivedrive
(𝐃𝐂 ) that
(DCrotates left moving
) that rotates the rightthe
left moving legright
con- leg
cerning the left one, at the same time the mass moves and stabilizes
concerning the left one, at the same time the mass moves and stabilizes the robot’s the robot’s centre
of gravity
centre of (see the example
gravity (see thecentre
example of gravity—COG—position)
centre of gravity—COG—position) within thewithin
left foot’s
the left
footprint;
foot’s footprint;
• • TheThe central driver
central (𝐃𝐂(D
driver ) stops;
C ) stops;
• • LeftLeft footfoot
(𝐃𝐋(D
) swivels
L ) swivels motor
motor isisstarted
startedto torotate the robot
rotate the robotaround
aroundthe theleft
left foot
foot by by
αL αangle;
L

• angle;
Left foot drive (DL ) is stopped when the final angular position is reached;
• • LeftCentral
foot drivecommon(𝐃𝐋 ) isdrive
stopped (DCwhen
) startsthetofinal angular
rotate position
right by ϕ angleis reached;
lowering the right leg
and at the same time the stabilizing mass moves to the upright position and the robot
statically stabilizes on both feet.
Then the sequence repeats. When the DC drive moves on ϕ angle, legs move in the
vertical direction by distance:
H = 2Rϕ, (2)
where: R is the radius of the gear and ϕ is the gear rotation angle.
During the walk, rotation of the foot is used for robot progression. The length of step S
depends on the value of the rotation angle α (αR or αL ) and the distance between the legs L:
α
S = 2Lsin (3)
2
where: L is the distance between the legs, and α foot rotation angle.
The rotational degree of freedom in the foot enables changing the direction of locomo-
tion, significantly exceeding the capabilities of conventional stepping robots.
Sensors 2022, 22, 4440 11 of 31

A three-dimensional printed prototype of the 3 DOFs robot [131] was driven by three
DC servos using the Pololu Maestro controller board. Thanks to the rotation of the foot, the
robot is characterized by unprecedented agility among bipedal robots. Extending RotoFoot
with the fourth degree of freedom (4 DOFs) at the torso, enabled the robot to climb stairs
by improving control of the COG location using an independent balancing drive (DG ) (see
Figure 1b). A prepared 3D printed prototype of 4 DOF can climb stairs [134]. The new type
of robotic synthetic drive can find different applications. For example, this robot can be
used as an intelligent robot-companion or delivery robot. This type of walking robot seems
to be particularly convenient to walk on flat surfaces; at the same time, the idea presented
also allows for climbing stairs.
A robot called LEONARDO (LEgsONboARDdrOne), or LEO for short, is a versatile
design that enables two main modes of movement walking as well as flying [137]. This
robot presents synchronized agile walking movements interspersed with flight manoeuvres.
This allows it to perform manoeuvres that are difficult for traditional walking robots such
as skating or walking on a rope. LEO consists of three main subsystems, namely a torso, a
propeller drive system, and two legs with pointing feet. The robot legs are constructed of
carbon fibre tubes and 3D printed carbon fibre reinforced joints with ball bearings. They
constitute a parallel kinematic mechanism with brushless DC (BLDC) drives with gears
located close to the torso, providing a compact form with reduced leg inertia. Both legs are
symmetrical, and each leg has three servos for actuation. The first is located at the pelvis
and moves the leg structure in the frontal plane of the LEO. The other two servo motors are
located at the front and the back of the hips and drive the parallel leg mechanism. Thanks to
applied solutions LEO weighs only 2.58 kg. Its height while walking is 75 cm. The solution
uses high-friction urethane polygamy as a pointed foot with a load cell for ground contact.
The robot uses sensors to detect the contact with the ground.
LEO can then execute the walking phase using inverted pendulum control, but this is
aided by the operation of the propellers. The robot’s construction allows it to change its
configuration during the flight phase. LEO can operate completely autonomously with its
onboard computers and sensor set; it also has other capabilities of moving all of which can
be found in [138].

4. Drive and Control Systems of Biped Walking Robots

Modern walking robots are equipped with drive control systems and sensor systems
to support both the operation of the robot’s mechanism and environmental recognition
systems. The solutions applied in these systems for main bipedal walking robots are
presented in this section.

4.1. Drive Systems

Ficht et al. [30] put forward the applied kinematic structure of the leg drive of modern
bipedal robots. Both rotary and linear drives are summarized. He distinguished between
five main causes of the structure of lower limb drives:
• Serial drive on-axis;
• Serial drive off-axis (requires gear);
• Parallel mechanism cranks-lever;
• Parallel mechanism, with a linear drive;
• Mixed serial/parallel mechanism.
Modern robots are driven by the use of electric drives, less often hydraulic ones [30].
Electric drive is used as:
• Harmonic drive (ASIMO [24], HRP-5P [121], Toro [105], TALOS [120];
• Cycloid driver (Cassie [118], Digit [122]);
• Muscle/tendons (Kengoro [116]);
• DC servo NimbRo-OP2 [119]);
• Serial elastic actuator SEA (Valkyrie [111], WALK-MAN [107]).
Sensors 2022, 22, 4440 12 of 31

Hydraulic drives are used as:

• Servo-valves (Atlas robots [101]);
• Electro-hydrostatic actuator—EHA (Hydra [117]).

4.2. Control Systems

Four possible criteria are commonly used as the equilibrium/stabilization indicators
for a bipedal robot: zero moment point (ZMP), point care map for limit-cycle walking,
angular momentum-based criterion, and footstep-based criterion. As a result, the first
two criteria are necessary to generate feasible biped motion, and the last two criteria are
possible assistive indicators that can be used to recover the biped balance caused by external
perturbation. There are explicit relationships between the ZMP, COG angular momentum,
and the footstep, see [6] for more details. However, there are three problems related to the
stabilization control of bipedal robots:
• First, how to generate the desired reference trajectory for a biped with high degrees
of freedom.
• The second is how to guarantee (feasibly balance/stabilized) reference trajectory for
the robot. This question is relevant if approximate models for trajectory planning
are used.
• The third is how to precisely track the desired angular joint references considering the
Sensors 2022, 22, 4440 13 of 32
computational complexity of the high degrees of freedom of the biped. For example,
Figure 2 suggests a general multi-level stabilization control for ZMP-based biped robot.

Figure
Figure 2.2.Multi-level
Multi-levelcontrol
controlarchitecture with
architecture four
with control
four layers:
control High-level
layers: control,
High-level Mid-level
control, con-
Mid-level
trol 1, Mid-level control 2, Tracking low-level control.
control 1, Mid-level control 2, Tracking low-level control.

There
There is
isno
noprevailing
prevailingapproach.
approach.Simple models
Simple cover
models basic
cover multisegmented
basic ballistic
multisegmented bal-
and
listic and passive gait models. Ballistic bipedal walking robots can analyse internalsuch
passive gait models. Ballistic bipedal walking robots can analyse internal signals sig-
as energy consumption or the level of disturbances [140–143]. Simple mutually coupled
Rayleigh oscillators for feedback control of a walking robot were described by de Pina et
al. [144] and Luo et al. [145]. Zielińska described four coupled oscillators generating real-
time outputs similar to human gait [57]. Chen et al. [146] presented sensor data fusion for
the state of body estimation in stable walking using feedback control. Klein and Lewis
Sensors 2022, 22, 4440 13 of 31

nals such as energy consumption or the level of disturbances [140–143]. Simple mutually
coupled Rayleigh oscillators for feedback control of a walking robot were described by
de Pina et al. [144] and Luo et al. [145]. Zielińska described four coupled oscillators gener-
ating real-time outputs similar to human gait [57]. Chen et al. [146] presented sensor data
fusion for the state of body estimation in stable walking using feedback control. Klein and
Lewis provided a neuro robotic model based on Golgi tendon organs, and spiking neural
networks [147]. Different feed forward strategies to recover from a trip or slip causing a fall
due to response in muscle(s) excitation status were described by Forner-Cordero et al. [148].
Neural control models coupled with dynamical models driven by joint moments showed
integrative properties of the neuromusculoskeletal systems within stable gait and indi-
vidual muscle contribution to trunk support [149]. Even gait disorders associated with
Guillan-Barré syndrome or spinal cord injury may be effectively reflected in walking assis-
tant robots [150,151]. Based on the previous studies, it can be assumed that stable bipedal
locomotion may be achieved by combining:
• Reflex-based control (e.g., artificial neural network) [152];
• Signals from local sensors;
• Simplification of bipedal robot kinematic construction;
• Adaptive compensation of small disturbances through controlling its dynamical
properties [153].
For more details on control strategies of biped walking that are based on energy
efficiency, see, e.g., [33,62,64]. Multi-level control architecture with four-level layers using
different sensors is presented in Figure 2.
High-level control. The highest level of the inverted pendulum problem has to be
solved using a model (IPM), see [6]. There are two essential problems at this control level:
• How to keep continuous COG state variables while changing the biped status/
orientation? A modification to the IPM is required as discussed in [2].
• How to reduce the modelling error caused by the IPM inaccuracy? This can be answered
by the mid-level control 1.
Mid-level control 1. It is responsible for a compensation of modelling error that results
from the high-level control. Proportional–integral–derivative (PID) control or advanced
control strategies, e.g., preview control can be used for regulation of the ZMP indirectly by
tracking the referential COG.
Mid-level control 2. It includes the inverse kinematics model of the robot and algo-
rithm based on the calculation of the desired COG generated by the last level control and the
referential foot trajectory. This control level can be avoided if task space coordinates-based
dynamics are used rather than joint space-based dynamics.
Tracking low-level control is based on the precise observation of the reference points
of the joints defined by inverse kinematics. A simple PID or advanced control structure can
be used in this control layer. This control is recommended to be distributed or decentralized
for a high number of DOFs biped mechanisms to avoid computational problems.

4.3. Sensor Systems

Position sensors on the motors provide data during locomotion that are key for
both current state identification during the gait and biped robot further steps planning.
Chen et al. [146] described a joint sensor system with a Kalman filter for the feedback
control of the walking robot.
The sensor system of biped robots was presented in [154–156]. Control of walking
robots heavily depends on the progress in the field of mechatronics and the development
of computer technology. Drive systems and advanced sensing and vision systems were
the key developments enabling stable control of walking robots. There are new solutions
for bipedal robots. The most advanced systems are developed in the USA and Japan, see
Table 1 for details. For a biped robot to walk safely in complex environments, several
sensor systems are required to sense/observe any obstacle or external objects. A good
Sensors 2022, 22, 4440 14 of 31

sensor system should be selected to recognize the correct foot placement while walking
with different configurations. The sensor system can be subdivided into five categories:
• Body orientation system. To capture the trunk tilt of the biped, inertial measurement unit
(IMU) sensors are commonly used, as they contain accelerometers and gyroscopes
that through sensor fusion can be used for reliable orientation estimation. They are
installed on the trunk in addition, some IMUs are placed on the feet to detect the feet
inclination. Incremental, high-resolution encoders are often connected to joint motor
shafts to measure joint positions and allow computation of positions and velocities.
For detailed characteristics of these sensors, see the examples [154–156].
• Foot sole sensor system (Force sensors). The ground reaction forces play an important
role in the stabilization of the biped mechanism and detecting ground stability. If these
forces are outside the stability region, the foot may slip, and the biped robot might be
not able to avoid a fall. Therefore, controlling these forces is necessary via confining
the ground reaction forces to stay within the support foot/feet. This strategy meets
the concept of ZMP. The ground reaction force wrench can be measured by placing
four six-axes force/torque sensors on the foot sole.
• Touch sensor system. Some biped robots are designed to work in a home environment
where there is a contact (touch) between the robot and human. Therefore, it is rec-
ommended to install the tactile sensor at specific locations to avoid trapping human
hands/fingers in-between the robot joints. For example, [156] has used 19 tactile
sensors placed inside the main elements of the robot. If these sensors are activated,
then the biped robot attempts to release the joint forces.
• Force sensing. In the case of electric drives, monitoring of forces that robots can apply
can be done using simple current sensing that can be part of the motor controller
or external circuitry. For hydraulic actuation systems, pressure sensors installed on
supply lines can be utilized to quantify force production.
• Audio sensor system. This sensor system is necessary for online communication with
humans where a multi-microphone system is built. For example, the solution [156]
installed seven audio sensors (microphones) on the head of the biped mechanism.
• Visual sensor system. Here, most typically, the head is equipped with a stereo camera-
based vision system to identify objects and avoid obstacles, see [157] for more details
on this topic.

4.4. Navigation Systems of Bipedal Robots in Uneven Terrain

Modern walking robots with environment recognition functions can move in un-
known terrain [158–162]. They appeared as a response to the modern threats, where robots
can replace humans in emergencies and harsh environments like fire, radiation (e.g., the
Fukushima disaster), or in outer space [159]. Robot environment recognition systems
are equipped with heads that include various sensors. The Light Detection and Ranging
(LIDAR) system is used for real-world orientation. Visual sensors are effectively utilized
for Visual Odometry (VO), and LIDAR sensors are used for LIDAR Odometry (LO) [162].
Unfortunately, such measurements commonly suffer from outliers in a dynamic environ-
ment, since frequently it is assumed that only the robot is in motion and the world is static.
Moving objects in the environment need to be filtered out to provide reliable odometry.
Developed by Carnegie Melon University Multisense SL/SLB, a camera-based vision
system was combined with a movable LIDAR sensor [159]. Such a system can be considered
a fully functioning humanoid robot head as it allows the robot to “see” a very large
scene around it without having to activate cameras or the neck. Several humanoid robots
participating in the DARPA Robotics Challenge (DRC) [159] are equipped with Multisense
SL/SLB, such as Atlas-DRC and Atlas-Unplugged [160], WALK-MAN [107]. Moreover, the
birdlike robot Digit [122] from Agility Robotics uses a LIDAR sensor.
Pieces of information from the camera-based system and LIDAR sensor are used by
special processing software. The authors [161] present an algorithm for the probabilistic
fusion of sensor data from different sensors (inertial, kinematic, and LIDAR) to produce
Sensors 2022, 22, 4440 15 of 31

a single consistent position estimate for a walking humanoid. Fallon et al. [160] describe
the perception and planning algorithms that have allowed a humanoid robot to use only
passive stereo imagery without LIDAR to safely plan footsteps to continuously walk over
rough and uneven surfaces without stopping. Experimental results confirmed this idea.
The robust Gaussian Error-State Kalman Filter for humanoid robot locomotion is presented
in [163]. The introduced method automatically detects and rejects outliers without relying
on any prior knowledge of measurement distributions or finely tuned thresholds.

4.5. Comparison of Known Modern Bipedal Robots Based on Human or Bird Walking
In Table 2, some features of the most known modern bipedal walking robots whose
design are based on human or bird’s gait are presented. Chosen robots are shown in order
of the year of their first introduction. Certainly, robots are systematically developed. In this
table summary of utilizing sensors, control systems, and type and number of drives are
also given. Additionally, for some robots, their walking speed and load-carrying capacity
for hands is also presented. All walking robots presented in Table 2 are equipped with
sensors for monitoring the position of the joints. In addition, some are also equipped with
torque or force sensors for joints. Most robots use IMU sensors for orientation in space.
A few use the LIDAR system for this purpose. All humanoid robots are equipped with
various types of vision systems. The Digit robot, which is rather a hybrid of a humanoid
robot (torso with arms) and bird base walking platform, also utilizes LIDAR. The Cassie
robot is not equipped with a camera. This model was mainly used to study walking and
running functions based on bird ideas, while environment recognition and processing were
not required in this project as the robot was controlled by an animator. The presented
comparison shows that not only LIDAR can produce a good surrounding recognition
function, but several robots also achieve excellent results in environment perception with
the sole use of cameras. Apart from the Atlas robot, all other robots presented in Table 2
are equipped with various types of electric drives. However, the Atlas robot [96–98,160],
which is equipped with a control system based on a LIDAR scanner and stereo vision and
utilizes hydraulic motors for driving joints (Table 2) presents excellent possibilities in terms
of precision and dynamics of movement, and the ability to move in difficult terrain.
An excellently tuned selection of sensors combined with state-of-the-art software
provides distinctive opportunities for human interaction. It should be emphasized that
the team of designers and programmers achieved these results in a relatively short time.
Compared to the other robots, Atlas is distinguished by a much lower weight owing to
the utilization of hydraulic drives that additionally provide excellent dynamics [103]. The
performance of the bipedal walking robots using the bird’s movement pattern is also
impressive: the low weight and the limited number of drives. Cassie [118] and Digit [122]
robots are characterized by the original construction of the legs using the cycloid drive,
which enables the use of very lightweight legs. It should also be emphasized that the
slightly smaller NimbRo-OP2 [117] robot is made with the 3D printing technique, which
ensures both low weight (due to the possibility of structure optimization) and a wide set of
options for developing and improving the elements of the structure.
Advanced walking robots, which are based on a humanoid gait model, require a
complex drive and control systems with many sensors. Moreover, a new type of knee-less
walking mechanism [135,136] requires an advanced control system. It is a result of the
number of DOFs. On the other hand, the developed synthetic walking mechanism for a
bipedal walking robot with swivel feet [131–134] does not require a complicated control
system due to the use of a small number of DOFs and its static stabilization method. The
3 DOFsRotoFoot robot requires three drives combined with three potentiometers-based
position sensors for closed-loop control ϕ, αR and αL α angles. The 4 DOFs robot, where
the extra DOF is used for an independent mass movement (Figure 1b), enabled climbing
stairs [134].
Sensors 2022, 22, 4440 16 of 31

Table 2. Comparison of known modern bipedal robots based on human or bird’s walking.

Height Elements of the Control System Type ** Speed

Robot Number
Manufacturer cm km/h
Year Joints IMU LIDAR Camera F/T of Drives
Mass, kg Load, kg
Walking robots based on human gait
HRP-5P 183 Stereo E. HD -
Japan Position x x 4x
2013 101 vision 37 13
Valkyrie 187 Position Multiple. E. SEA -
NASA 7x 2x
2013 129 Torque cameras 44 -
Toro GAC 174 Position RGB&D E. HD 1.8
2x
2014 Germany 76 Torque camera 39 10
Atlas–N. G. Boston 150 Position Stereo H. S-v.
x 5.4
2015 Dynamics 75 Force vision 30
WALK-MAN IIT 191/185 Position Multiple E. SEA -
2x 2x
2015 (2018) Italy 132/102 Torque cameras 29 10
Kengoro Tokyo 167 Position Stereo E. Muscle -
x 2x
2016 University 56.9 Tension vision Tend./106 -
NimbRo-OP2 Bonn 135 Stereo E. DCSM -
Position x
2017 University 19 vision 34 -
TALOS PAL 175 Position RGB&D E. HD -
x
2017 Spain 95 Torque camera 32 -
Walking robot based on birds gait
Cassie * Agility 115 E. CD
Position x 5
2016 Robotics 31 10
Digit Agility 155 4x Depth E. CD -
Position x x
2019 Robotics 42.2 camera 16 -
* Robot with only lower body. ** Drive type: E Electric, CD—Cycloid Drive, DCSM—DC servo Motors,
HD—Harmonic Drive, SEA—Series Elastic Actuator., H.—Hydraulic, S-V.—Servo-Valves.

4.6. Energy Sources of Robots

Robots can be actuated in a number of ways; hydraulic, pneumatic, and electric
drive systems are common solutions. In any case, the control and sensing systems require
electricity for operation. This implies need for electrical supply in any way, which can be
achieved by providing separate electric energy source for control or by utilizing energy
conversion from the main energy source. Yang et al. [164] presents the summary of different
energy sources for robotics. He divides all sources on:
• Energy storage, including batteries and capacitors/supercapacitors;
• Power generators—fuel cells, classical electromagnetic generators, and solar cells;
• Power harvesting (phototovoltaic, electrochemical, wireless, thermoelectric, etc.) and nano-
generators (micro-/nano-energy sources, self-powered sensors, and flexible transducers).
The main energy source for mobile robots is rechargeable batteries. For best perfor-
mance, low weight, high current draw capability, and high capacity are required. Low-
est cost per Wh can be obtained from lead-acid batteries, but they provide the lowest
energy density per weight unit (35–40 Wh/kg) [165]. On the other end, LiFePO4 batter-
ies are placed with the highest unit cost but also highest energy density per unit weight
(90–160 Wh/kg) [166]. Some years ago, rechargeable battery technologies were used (Ni-Cd
or Ni-MH–cell voltage 1.2 V), but the best are now Li-ion batteries with the highest energy
density 100–265 Wh/kg, cells voltage 3.6 V, without memory effect and low self-discharge
effect (1.5–2% per month) [167]. A problem with Li-ion batteries is the possibility to fire.
Li-ion batteries for walking robots (1.1 kWh) produced by CEO [168] uses 4 ventilators
for cooling of 98 cell (302.4 V). Detailed review for battery technology is provided in [169].
Li-ion batteries, however, have two main limitations—low energy capacity as compared to
combustible fuels (petrol for example provides 13 kWh/kg, natural gas—15 kWh/kg, and
hydrogen—34 kWh/kg), and relatively long charging time as compared to refuelling time.
Compressed air can be stored with energy density in the range of 398 Wh/kg, which places
it somewhere between batteries and combustion engines or fuel cells [170]. Hydrogen fuel
cells are especially interesting, as electrical energy production is achieved without any
moving components and almost noiselessly. Drawback of fuel cell technology is the weight
Sensors 2022, 22, 4440 17 of 31

of hydrogen tanks [171]. Diesel motor combined with hydraulic pumps and/or generator
is a relatively simple solution for large robots; in smaller robots, such construction would
have a significant vibration problem [172]. Fluid powered actuators provide advantage
over electrical drives in terms of lower weight and high force, low speed characteristics that
in electric drives are achieved using gearboxes that add to system weight and complexity.
Fluid actuators characteristics are especially desired in walking robots.
Robots designed for short time operation can be powered by compressed air, electricity
from supercapacitors, or batteries offering a relatively simple design. Electrical power
storage is important in any case due to the control electronics and sensor systems utilized in
robotics. Energy recovery or harvesting during operation requires an intermediate energy
storage system that can accept large energy input in short time—for example in braking
energy recovery. For this purpose, in case of electrical energy, supercapacitors are a good
alternative; in case of fluid powered robots, pressure tanks serve the same function. Me-
chanical energy can also be stored for example in flywheels [173] or springs. Nevertheless,
energy conversion between different systems is one of the sources of inefficiency of the
drive train; therefore, it is foreseen that in the future most robots are going to utilize an
energy source mix that supports actuation and control specific to the work environment
of the robot and special application conditions. Currently, there is no ideal solution that
would work in all applications. For indoor applications, combustion engines are not well
suited due to noise and emissions, but recharging cycles can be planned with proper dock-
ing stations offering either pressurized gas, electricity, or hydrogen supply. For outdoor
applications, and long operational time requirements, energy harvesting will be utilized as
range extender, but the main power source will most likely become hydrogen or biofuels
due to practical reasons and cost.
The robot size has a significant impact on energy source selection. Small scale with
short duty cycles robots can be operated with just energy harvesting mechanisms and
simple electric actuation systems that offer one energy source for all components is a clear
winner, cost for dense electrical energy storage in small scale can also be justified in many
applications. As the size of the robot increases and the duty cycle lengthens, denser energy
storage with lower costs per unit energy is needed or fast energy refuelling either by battery
swapping, tank refill, or quick charging mechanisms is required. The larger the size of the
robot, the more benefit can be gained from energy source diversification and optimization
for individual active robotic components.
For each robot, designers solved problems of energy source. DURUS, an 80 kg robot,
used onboard 2.2 kWh lithium-polymer battery [103]. The BigDog robot, for instance,
utilizes a small go-cart petrol motor (11 kW) for driving hydraulic pump and electric
generator [174]; nevertheless, it was considered too noisy for military applications and
development was turned towards fully electric drive. Pneumatic actuation can be achieved
by integrating a compressor that is powered by electric or combustion motor, or by utilizing
an air tank for compressed air storage. Robot Lucy is an example where external compressed
air tanks or compressors are utilized [19].
Robots driven using hydraulic energy: PETMAN [89], ATLAS, and ATLAS DRC [102]
used external sources of supply pressure. ATLAS unplugged [102] used source placed on
board what increased mass from 152 kg to 182 kg. Actual version of ATLAS [103] based on
improved design of hydraulic system has, similar to humans, a mass of 80 kg with a height
of 151 cm.
Thangavelautham et al. [175] presents simulation results for a HOAP 2 humanoid robot
that suggests a fuel cell powered hybrid power supply, superior to conventional batteries.
Possibilities to use new energy sources present a 88 mg micro robot RoBeetle [176]
powered by catalytic combustion of methanol, whose specific energy is 5.6 kWh/kg. For
comparison, the specific energy of an animal fat 10.6 kWh/kg and a good Li-ion battery is
only 0.5 kWh/kg.
Intensive work is being carried out on the use of photovoltaics to power robots [177].
The relatively small surface area available for solar cells on the robot cannot convert enough
Sensors 2022, 22, 4440 18 of 31

energy to keep the power-hungry robot functioning, but with the technology, the cheaper
perovskite cell achieves almost 24% efficiency, and specialist designs like the multi-junction
(MJ) solar cell achieves as much as 40% efficiency [177]. Successful solutions of a solar robot,
shaped like a sphere, are presented [178] and with wheels, Tertill [179] or Vitirover [180].
There is also a known small solar walking robot [181]. It is possible to use solar power to
charge of rechargeable battery pack humanoid robots such as ATLAS [177].
The intensive development of batteries related to the automotive industry [182] allows
us to expect that these sources will become the basic power supply for robots both in terms
of propulsion power andin hybrid systems and control. The main directions are:
• Improve system of management of conventional Li-ion batteries;
• Improved design of Li-ion batteries by changing construction and composition lowers
costs and improves performance (cobalt-free lithium-ion battery, mesoporous silicon
microparticles and carbon nanotubes, lithium-sulphur batteries);
• New design and chemistry of batteries improve performance (vertically aligned carbon
nanotube (VACNT), aluminium-air light battery);
• Design which improved time of charge (solid state lithium-ion batteries with sulphide
superionic conductors);
• Structural batteries using carbon fibres as the negative electrode while the positive is a
lithium iron phosphate; the latest battery has a stiffness of 25 GPa. It is possible to use
it to design superlight electric vehicles and also walking robots.
This promises that future batteries in a short time will be lighter and will achieve
much greater capacity, which will provide walking robots with a long operating time.

4.7. Energy Efficiency of the Walking Robot’s Motion

In some papers, the authors deal with the energy efficiency of a walking robot. This
issue is very important due to the fact that robots are equipped with a limited amount of
energy stored in the power source. This resource determines the range of the robot and
depends on the efficiency of the drive system. The issue of energy efficiency of moving
humans and animals was for the first time analysed by Trucker [183], who introduced the
concept of the cost of transport (COT):
P
COT = (4)
Wv
where: P—the power of move, W—weight, v—speed. Since the weight of the robot depends
on the mass and gravitational parameter, the cost can be also presented as follows [127]:
P
COT = (5)
Mgv

where: P—Power of robot, M—mass of the robot, v—speed, g—standard gravitation.

Kajita et al. [184] and Sakagami et al. [185] proposed specific resistance (SR) as an
index for evaluating the energy efficiency of a mobile robot. It is based on the total energy
consumption of d distance to the gravitational potential energy:
E
SR = (6)
Mgd

where: E—Energy consumption, M—mass of the robot, d—distance of the move.

SR parameter is equivalent to the COT parameter. If the parameters of (4), (5), or (6)
relations are all expressed in a consistent system of units, the quantity is dimensionless
and has the same value in any system of units. Kashiri et al. presented an overview of
principles for energy-efficient robot locomotion [186]. Based on references, it is possible to
obtain some results of the COT or SR value of different robots and also to compare humans
and four-legged robots. For MIT’s cheetah quadruped, which exploits electrical energy
regeneration/recycling, the COT is about 0.5. A human is down around 0.2 [113], while
Asimo (54 kg) exhibits COT = 2 for 1.5 m/s [185]. In the DARPA competition between
Sensors 2022, 22, 4440 19 of 31

robots, the best was SRI’s DURUS robot [113,187], which, with a fully charged battery,
walked a distance of 2.05 km to run the battery dry in 2 h, 35 min, 43 s. The walking of the
DURUS robot can be seen in the movie [114]. While walking, it used around 350 watts of
power, giving it an average COT of about 1.5, which is better than the Atlas robot, which
has a COT of 20 [113]. SRI estimates that in the future, it will be possible to obtain for the
DURUS robot a COT under 1, and with its onboard 2.2 kWh lithium-polymer battery, the
80 kg robot should be able to walk 10 km [113].
SANDIA laboratory’s present special design of Walking Anthropomorphic Novelly
Driven Efficient Robot for Emergency Response (WANDERER) [123]. It is possible to see
the mechanism of the robot and its walking [124]. One of the best bipedal walking robots
in terms of transportation cost is Cassie [118,188]. The new design of the Cassie robot
(31 kg) has a low COT value, allowing Cassie to run for 6–8 h on a single charge. This
robot based on the original design obtains excellent running results, completing 5 km in
about 43 min [189]. When walking at 1.0m/s, using a total of 200 watts of power, while
performing different locomotion behaviours such as squatting, the calculated COT was 0.7.
For the LEO robot, v = 0.2 m/s COT = 108 was determined during walking [137]. When
walking on the ground, LEO sucks down 544 watts, of which 445 watts go to the propellers
and 99 watts are used by the electronics and legs [137]. LEO has a nominal walking speed
of 20 cm/sec, and its total velocity relative to the ground can undulate significantly using
intermittent near-ground flight. When flying, LEO’s power consumption almost doubles; it
can also fly with speeds of 1–3 m/s. This high value is due to the high energy expenditure
of the propeller drive system to stabilize the robot during gait. When flying at v = 1 m/s,
COT = 48, and at v = 3 m/s, COT decreases to 15.5 [137].

5. Future of Bipedal Walking Robots

The future development of walking robots will depend on both the improvement and
the appearance of new constructions, drives, and the use of various sensors for control
movement systems as well as interactions with the environment and humans. Certainly, it
will require the improvement of the operating software.

5.1. Limitations of Development of Walking Robots

The main limitations thatcan be deduced from the previous studies can be summarized
as below:
• Theoretical limitations—there is a lack of comprehensive theories regarding the biome-
chanics, kinematics, and mechanisms of control/coordinating gait, ambulation, and
clinical gait assessment; thus, it is hard to understand all aspects of bipedal gait, both
physiological and pathological, and reflect them within the bipedal walking robot.
• Technical limitations—current technologies of bipedal walking robots are:
# hard to assess due to limitations of computational models;
# complicated design and control;
# cost-effective only in several specialized applications, e.g., space exploration,
military purposes, etc.
• Cognitive and ethical limitations—despite wide development (robots for the elderly,
robotic toys for children, etc.), there is a need to increase common attention to the
ethical limitations of using technology (including ICT and AI) for care interventions
for people with limited self-awareness, insight, and orientation.

5.2. Achievements and Opportunities

There is a clear increase in open-source walking robotics projects in the last decade [130].
This indicates that the walking mechanisms will be widely used by students and even
children, at first as toys, and later as scientific projects. Awareness of robotic capabilities
in the general public is important before assistive robots and cooperation robots become
easily available and widespread. Knowing from experience how robots perceive their
surroundings and how to interact with them safely will lower psychological barriers that
Sensors 2022, 22, 4440 20 of 31

people have against new inventions. This in turn should lead to market-driven research on
walking robotics directly applicable to common everyday tasks.
Bipedal walking robots may improve our understanding of the underlying neuronal
control as far as biomechanical principles, mechanics, and muscle functioning of gait in
individuals with gait impairments [147,149]. Modern bipedal robots are based on nat-
ural structures that move perfectly, both based on human gait patterns (ATLAS [103],
Valkyrie [111], etc.) or birds’ template (Cassie [118], Digit [122]). Owing to the advanced
sensory systems, precise, fast, and efficient drives can perfectly cope with moving in
unknown terrain and respond to disturbances to maintain stability.
An important direction of development is the use of bipedal robot motion kinematics
based on bird gait and run. The Cassie robot undertook testing in 2021 (with operator
supervision) [189] and travelled 5 km in around 44 min at an average speed of 6.81 km/h.
In one lap, the robot reached 7.74 km/h. The maximum speed achieved by this robot on
long distances is more than the Atlas robot could achieve, as it reaches 5.4 km/h. Note,
however, that the Atlas uses a hydraulic drive, while the Cassie robot is powered by a DC
motor system. A bipedal walking robot became the basis for developing a neuromechani-
cal simplified planar musculoskeletal model of human lower body biomechanics with a
controller based on a dynamic artificial neural network with central pattern generators
(CPGs) coupled with force and motion feedback to generate the appropriate muscle forces
needed for walking. Separate neural networks generate the rhythm and create the gait
pattern, especially stable in the sagittal plane without inertial sensors, a centralized posture
controller or a walker in a manner similar to human walking (speed 0.850–1.289 m/s with
a leg length of 0.84 m, also on 5◦ slopes without additional controller actions) [190]. An
underwater bipedal walking soft robot based on a coconut octopus was designed and a
machine vision algorithm was used to extract motion information for analysis—such a
walking robot can achieve an average speed of 6.48 cm/s [191]. The bipedal walking robot
has also become a test case for the use of shape memory alloy (SMA) springs as artificial
leg muscles [192]. This is important for, among other things, locomotion rehabilitation and
the development of assistive devices [193]. A better understanding of human movement
increases the possibility of successfully combining humans and technology.
In addition to the construction of bipedal robots based on nature bipeds (humans or
birds), the subject of research is unconventional solutions for bipedal robots:
• Innovative walking robots with swivel feet [131–134] with very simple kinematics
provide excellent manoeuvrability (3 DOFs) and also the ability to climb stairs in the
4 DOFs version;
• A two-legged robot called SLIDER [135,136], which uses a sliding joint in each leg to
replace the knee–hip rotary motion that is used in nature for leg lifting;
• The bipedal robot LEO with a versatile design that enables two main modes of move-
ment by walking as well as flight [137,138].
Especially the latter solution seems to be useful due to the possibility of the robot
moving both above the surface and in the air. Hence, the design and control of the
multimodal robotic locomotion LEO allows for rope walking or skateboarding, which were
previously challenges for bipedal robots [138].

5.3. Directions of Future Research

The aim of further work should be to combine the kinematics possibilities of bipedal
robots and the possibility of safe interaction with humans. The authors of [158] presented
a literature review on sensors of humanoid robot heads, especially focusing on aspects
related to human–robot interaction. It has been observed that the robotic vision is the most
widespread perception sensor system utilized in humanoid robots. This is mostly due
to the amount of information that vision provides about the surroundings. The authors
determined that there are two main types of humanoid robot heads and proposed a division
into robot heads with non-expressive faces and robot heads with expressive faces. The
latter is more human-like and, therefore, often perceived as more friendly. This will be
Sensors 2022, 22, 4440 21 of 31

the direction of further work along with the development of artificial intelligence, hybrid
intelligence, and embodied intelligence of robots [194,195].
Taking into consideration wider perspective directions for further research should
focus on:
• from a scientific point of view: knowledge sharing, including open-source solutions;
• from a technological point of view: on the development of robot navigation and
artificial intelligence systems;
• from an organizational point of view: interdisciplinary collaboration among various
research centres, virtual research teams, platforms for experiences, knowledge, and
project sharing;
• from a clinical point of view: taking into consideration advanced applications of the
aforementioned solutions in everyday therapy;
• from a societal and industrial point of view: dissemination of the knowledge and
experiences, building social awareness concerning wider use of the bipedal robot
walking in various areas of the daily life.
The most important research goals for the next several years may cover:
• high-quality studies to address research gaps within neural control and biomechanics
of bipedal gait;
• the use of a highly dynamic hydraulic drive and the use of an innovative sensors
system based on the LIDAR scanner system combined with an artificial vision system
using stereo cameras set a completely new level in the field of robot’s world analysis;
• space exploration plans will mostly benefit from the development of autonomous
walking robotics [194];
• a very important direction of the development of the bipedal walking robots mech-
anism will be further efforts to find solutions with high energy efficiency (low COT
value). The goal is to approach and maybe even overcome the limit COT = 0.2 defined
for some human walking conditions. This will allow increasing the range of robots
with limited battery capacity.

5.4. Main Direction of Walking Robot Applications

The work carried out in the field of robot kinematics has produced valuable results
indicating the possibilities of developing robot applications in various fields. Now walking
robots are able to move both in urban terrain with unified flat wall and floor surfaces
and stairs. Today’s robots are able to move in terrain with unfamiliar configurations. The
development of energy sources and the improvement of energy efficiency point to the
possibility of covering considerable distances. Some robots equipped with hands and
touch sensors and equipped with a vision and recognition system can manipulate objects
using cooperation of the upper limbs (Atlas [103], Digit [112]). Robots are also capable
of using tools. In parallel with the development of the design and kinematics of robots,
the development of artificial intelligence of robots is progressing. These developments
are taking place at a pace beyond recent imaginings. Today’s robots can soon replace
humans in dangerous and tedious tasks. Robots are being developed for tasks in space
(Valkyrie [111]). Work is being carried out also on the use of walking robots for military
purposes. The use of robots is likely to be driven by the need to increase productivity and
save the environment.
Additionally, research on bipedal walking robots contributes to the scientific under-
standing of human movement, including the transition from walking to running [195]. The
development of assisted locomotion with the use of exoskeletons may eventually make
it possible to do away with wheelchairs or ramps, and disabled people will be able to
enter places where fully able-bodied people enter. Bipedal locomotion and upright body
position are also necessary for proper functioning of the human body (e.g., bladder and
bowel function—certain muscles have to tense up and to relax).
Sensors 2022, 22, 4440 22 of 31

Research into bipedal walking robots, both towards improving kinematic capabilities
and developing artificial intelligence, could spur the creation of further well-designed
assistive technologies:
• Solving problems in uncertain environments and assistance in hazardous places—
firefighter support, zones with radioactive contamination, extreme temperatures,
explosion risk zones, mines, outer space, etc.
• In the future, intelligent bipedal robots can be used in factories to serve as a replace-
ment or a collaborator for human workers;in services for jobs that require taking on
strenuous, uncomfortable positions or low-paying jobs that humans do not want to do,
interactive robots may open new chances toward human–robot relationships, social
awareness, activity monitoring, activity eliciting, and learning.
• Artificial environments such as virtual reality and augmented reality and brain–computer
interface (BCI)may be more advanced alternatives for the traditional human–robot
interface, providing multimodal interaction comparable to inter-human communication.
• Mobile technologies open new possibilities of remote control, e.g., for children/elderly
safety and telemedicine/telerehabilitation purposes, as well as a general condition,
sport, and leisure activities.
• Application of intelligent robots in space missions, both on their own and in supporting
people, which may require the ability to walk.
• In conclusion, it should be stated that we are approaching the moment when robots in
a much more perfect form than so far will enter our lives on two legs equipped with
human-friendly artificial intelligence.

6. Conclusions
The presented survey of the literature as well as the previous work of the authors
allows the formulation of the following conclusions:
• The artificial gait solutions of legged robots are based on the essence of the biolog-
ical standard that has been created in the natural process of evolution through the
development of joint muscles and the excellent biological control system;
• Gait inspirations for versatile application of walking robots can be gained from birds,
insects, or other animals, not popularly used in walking research currently;
• Characterizing the essence of human gait and achieving it inmechatronics, drives, and
control systems enabled the creation of solutions imitating this way of locomotion;
• The artificial gait systems based on a human biological pattern are built with varying
degrees of complexity from 12–30 DOFs (close to human kinematics) to 6 DOFs or less
(extremely simplified);
• Models imitating human gait mainly use electric drives with various motor types; in
addition, especially recently developed walkers provide excellent performance using
hydraulic drives or artificial muscles as in the Atlas robot;
• The new advanced design of bipedal robots based on natural patterns are Cassie and
Digit, which use the pattern of bird gait. These robots use, with success, electric drives
and are the fastest bipedal walking robots nowadays;
• The newly proposed solution based on the unique (synthetic) idea is a walking robot
without a knee (10 DOFs);
• The bipedal innovative walking mechanism with excellent manoeuvrability equipped
with two legs sliding in body with swivel feet allow to move on flat terrain at 3 DOFs,
and at 4 DOFs, also climbing stairs becomes possible;
• An interesting design of walking robots is the robot Leo with a versatile drive, which
can gait on two legs as a bipedal robot and fly as a drone;
• The conventional sensor system of advanced solution biped walking robots needs
five categories of sensors systems of body orientation, foot sole, force, touch, vision,
and audio.These kinds of sensors need both to solve the problem of moving robots by
walking and communication of robots with the environment;
Sensors 2022, 22, 4440 23 of 31

• Modern bipedal walking robots allow movement on flat surfaces in the field, the best
solutions of a bipedal robot equipped with visual and LIDAR sensors can be used in
uneven terrain to solve problems in conditions dangerous for humans—fire hazards,
radioactive contamination hazards, or in outer space;
• Advanced walking robots that are based on natural biped gait need complicated drive
and control systems with many sensors. This is related to the stabilization of the
inverted pendulum system created by the legs. This problem does not exist in the
presented biped (3 DOFs or 4 DOFs) walking robots with swivel feet;
• Combining the possibilities of building bipedal walking robots and equipping them
with AI and human communication systems such as speech synthesis and recognition,
and affective computing systems open new applications for these robots;
• The running time of the robots depends both on the performance of the power supply
system but also on the COT parameter, which reaches for the DURUS robot a value
of 1.5, and for the CASSIE robot COT = 0.7, when for humans the observed value is
COT = 0.2;
• Walking robots use both hydraulic drives (ATLAS), which provide the greatest dy-
namics, and electric drives (DURUS, CASSIE, etc.). Rapid development of powerful
batteries suitable for fast recharging is foreseen.
Modern robots have reached an excellent level of construction design, and drive sys-
tems that are capable to mimic perfectly natural walking, including automated locomotion
in unknown terrain. It is especially important that the robots should become user-friendly
and would not be in any way dangerous. The military applications of biped robotics are
somewhat disturbing and might create a physical barrier to introduce biped robots into
civil applications.
Future research will focus on the development of robot navigation and intelligence
systems, as well as the interfaces for communication and collaboration with humans.
Taking into consideration a wider perspective, directions for further research should
focus on:
• from a scientific point of view: knowledge sharing, including open-source solutions;
• from a technological point of view: on the development of robot navigation and
artificial intelligence systems;
• from an organizational point of view: interdisciplinary collaboration among various
research centres, virtual research teams, platforms for experiences, knowledge, and
project sharing;
• from a clinical point of view: taking into consideration advanced applications of the
aforementioned solutions in everyday therapy;
• from a societal and industrial point of view: dissemination of the knowledge and
experiences, building social awareness concerning wider use of the bipedal robot
walking in various areas of the daily life.
As authors represent nine independent research centres, each centre works on its
project to solve part of the larger problem. We have developed collaboration mechanisms
for remote group work utilizing online meeting platforms and cloud document sharing to
facilitate long-distance teamwork. Information exchange and synchronization in terms of
research funding applications are one of the keys to organizing distributed expert groups
for multidisciplinary problems.
Sensors 2022, 22, 4440 24 of 31

Author Contributions: Conceptualization, T.M. (Tadeusz Mikolajczyk), E.M., H.F.N.A.-S. and T.M.
(Tomasz Malinowski); Methodology, T.M. (Tadeusz Mikolajczyk), E.M. and H.F.N.A.-S. Investigation,
T.M. (Tadeusz Mikolajczyk), E.M., H.F.N.A.-S. and T.M. (Tomasz Malinowski); Resources, T.M.
(Tadeusz Mikolajczyk), E.M., H.F.N.A.-S. and A.K.; Writing-Original Draft Preparation T.M. (Tadeusz
Mikolajczyk), E.M., H.F.N.A.-S. and A.K.; Writing-Review & Editing, T.M. (Tadeusz Mikolajczyk),
E.M., H.F.N.A.-S., T.M. (Tomasz Malinowski), A.K., D.Y.P., T.P., F.H., K.G., D.M., M.M. and F.H.;
Visualization, T.M. (Tadeusz Mikolajczyk), H.F.N.A.-S., T.M. (Tomasz Malinowski), A.K. and T.P.;
Supervision, T.M. (Tadeusz Mikolajczyk), E.M., H.F.N.A.-S., T.M. (Tomasz Malinowski), A.K., D.Y.P.,
T.P., F.H., K.G., D.M., M.M. and F.H. All authors have read and agreed to the published version of
the manuscript.
Funding: Part of this work was conducted as a part of work within a project “NeuroPerCog: develop-
ment of phonematic hearing and working memory in infants and children (head: Prof. Włodzisław-
Duch)”. The project is funded by the Polish National Science Centre (DEC-2013/08/W/HS6/00333,
Symfonia 1).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations

AI artificial intelligence
ASD autism spectrum disorder
BBWR bird biped walking robot
BCI brain-computer interface
BLDC brushless DC
COG centre of gravity
COT cost of transport
CPG central pattern generator
DC central drive of legs move
DG independent balancing drive
DL drive of left foot
DR drive of right foot
DC direct current
DOFs degrees of freedom
DSP double support phase
EHA electro-hydrostatic actuator
HBWR human biped walking robots
ICT information and communications technology
IMU inertial measurement unit
LIDAR light detection and ranging
PID proportional–integral–derivative
SBWR synthetic biped walking robots
SMA shape memory alloys
SR specific resistance
SSP single support phase
VO visual odometry
ZMP zero moment point

References
1. Morecki, A.; Knapczyk, J. CISM courses lecture notes. In Basis of Robotics: Theory and Components of Manipulators and Robots;
Springer: New York, NY, USA, 1999; Volume 402.
2. Wolovich, W.A. Robotics: Basic Analysis and Design; Saunders College Publishing: Philadelphia, PA, USA, 1987; Volume 393.
3. Lima, P.; Ribeiro, M.I. Mobile Robotics; Course Handouts; Instituto Superior Técnico/Instituto de Sistemas e Robótica: Lisboa,
Portugal, 2002.
Sensors 2022, 22, 4440 25 of 31

4. Silva, M.F.; Machado, J.T. A literature review on the optimization of legged robots. J. Vib. Control 2012, 18, 1753–1767. [CrossRef]
5. Vaughan, C.L. Theories of bipedal walking: An odyssey. J. Biomech. 2003, 36, 513–523. [CrossRef]
6. Bekey, G.A. Autonomous Robots: From Biological Inspiration to Implementation and Control; MIT Press: Cambridge, MA, USA, 2005;
p. 577; ISBN 0-262-02578-7.
7. Raibert, M.H. Legged robots that balance. IEEE Expert 1986, 1, 89. [CrossRef]
8. Al-Shuka, H.F.N.; Corves, B.; Zhu, W.H.; Vanderborght, B. Multi-level control of zero-moment point-based humanoid biped
robots: A review. Robotica 2016, 34, 2440–2466. [CrossRef]
9. Al-Shuka, H.F.N.; Allmendinger, F.; Corves, B.; Zhu, W.H. Modeling, stability and walking pattern generators of biped robots:
A review. Robotica 2014, 32, 907–934. [CrossRef]
10. Meghdari, A.; Shariati, A.; Alemi, M.; Nobaveh, A.A.; Khamooshi, M.; Mozaffari, B. Design performance characteristics of a social
robot companion arash for pediatric hospitals. Int. J. Hum. Robot. 2018, 15, 1850019. [CrossRef]
11. Chen, K.; Lou, V.W.; Tan, K.C.; Wai, M.; Chan, L. Effects of a humanoid companion robot on dementia symptoms and caregiver
distress for residents in long-term care. J. Am. Med. Dir. Assoc. 2020, 21, 1724–1728. [CrossRef]
12. Qidwai, U.; Kashem, S.B.A.; Conor, O. Humanoid robot as a teacher’s assistant: Helping children with autism to learn social and
academic skills. J. Intell. Robot. Syst. 2020, 98, 759–770. [CrossRef]
13. Robinson, P.; el Kaliouby, R. Computation of emotions in man and machines. Phil. Trans. R. Soc. B 2009, 364, 3441–3447. [CrossRef]
14. Vukobratović, M.; Borovac, B. Zero-moment point—Thirty five years of its life. Int. J. Hum. Robot. 2004, 1, 157–173. [CrossRef]
15. Kim, D.; Seo, S.-J.; Park, G.T. Zero-moment point trajectory modelling of a biped walking robot using an adaptive neuro-fuzzy
system. IEEE Proc. Control Theory Appl. 2005, 152, 411–426. [CrossRef]
16. Park, I.-W.; Kimand, J.-Y.; Oh, J.-H. Online biped walking pattern generation for humanoid robot KHR-3 (kaist humanoid
robot-3: Hubo). In Proceedings of the 2006 6th IEEE-RAS International Conference on Humanoid Robots, Genova, Italy,
4–6 December 2006; pp. 398–403. [CrossRef]
17. Zhu, W.-H. Virtual Decomposition Control: Toward Hyper Degrees of Freedom Robots; Springer Science & Business Media:
Berlin/Heildelberg, Germany, 2010; Volume 60.
18. Golliday, C.; Hemami, H. An approach to analyzing biped locomotion dynamics and designing robot locomotion controls. IEEE
Trans. Autom. Control 1977, 22, 963–972. [CrossRef]
19. Vanderborght, B.; Van Ham, R.; Verrelst, B.; Van Damme, M.; Lefeber, D. Overview of the Lucy project: Dynamic stabilization of a
biped powered by pneumatic artificial muscles. Adv. Robot. 2008, 22, 1027–1051. [CrossRef]
20. Zielińska, T. Walking Machines: Basics, Design, Control and Biological Patterns; WydawnictwoNaukowe PWN: Warsaw, Poland, 2013.
21. Raibert, M.; Tzafestas, S.; Tzafestas, C. Comparative simulation study of three control techniques applied to a biped robot.
In Proceedings of the IEEE Systems Man and Cybernetics Conference—SMC, Le Toruquet, France, 17–20 October 1993; Volume 1,
pp. 494–502. [CrossRef]
22. Vukobratovic, M.; Ciricand, V.; Hristic, D. Contribution to the study of active exoskeletons. In Proceedings of the 5th International
Federation of Automatic Control Congress, Paris, France, 12–17 June 1972; pp. 13–19.
23. Vukobratovic, M. When were active exoskeletons actually born? Int. J. Hum. Robot. 2007, 4, 459–486. [CrossRef]
24. Hirose, M.; Ogawa, K. Honda humanoid robots development. Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 2007, 365, 11–19.
[CrossRef] [PubMed]
25. Wahde, M.; Pettersson, J. A brief review of bipedal robotics research. In Proceedings of the 8th UK Mechatronics Forum
International Conference (Mechatronics 2002), Enschede, The Netherlands, 24–26 June 2002; pp. 480–488.
26. Bezerra, C.A.D.; Zampieri, D.E. Biped robots: The state of art. In International Symposium on History of Machines and Mechanisms;
Ceccarelli, M., Ed.; Springer: Dordrecht, The Netherlands, 2004. [CrossRef]
27. Shi, X.; Gao, J.; Lu, Y.; Tian, D.; Liu, Y. Biped Walking Based on Stiffness Optimization and Hierarchical Quadratic Programming.
Sensors 2021, 21, 1696. [CrossRef]
28. Ficht, G.; Farazi, H.; Brandenburger, A.; Rodriguez, D.; Pavlichenko, D.; Allgeuer, P.; Hosseini, M.; Behnke, S. NimbRo-OP2X:
Adult-sized open-source 3D printed humanoid robot. In Proceedings of the 2018 IEEE-RAS 18th International Conference on
Humanoid Robots (Humanoids), Beijing, China, 6–9 November 2018; IEEE: Piscataway, NJ, USA, 2019; pp. 1–9. [CrossRef]
29. Ye, X.; Bin, L.; Anhuan, X.; Dan, Z. A review: Robust locomotion for biped humanoid robots. In Proceedings of the Journal of
Physics: Conference Series, 2020 4th International Conference on Control Engineering and Artificial Intelligence (CCEAI 2020),
Singapore, 17–19 January 2019; Volume 1487. [CrossRef]
30. Ficht, G.; Behnke, S. Bipedal humanoid hardware design: A technology review. Curr Robot. Rep. 2021, 2, 201–210. [CrossRef]
31. Gubta, S.; Kumar, A. A brief review of dynamics and control of underactuated biped robots. Adv. Robot. 2017, 31, 607–623.
[CrossRef]
32. Westervelt, E.R.; Grizzle, J.W.; Chevallereau, C.; Choi, J.H.; Morris, B. Feedback Control of Dynamic Bipedal Robot Locomotion; CRC
Press: Boca Raton, FL, USA, 2007.
33. Sadati, N.; Dumont, G.A.; Hamed, K.A.; Gruver, W.A. Hybrid Control and Motion Planning of Dynamical Legged Locomotion; IEEE
Press: Piscataway, NJ, USA, 2012.
34. Milton, J.G. Introduction to focus issue: Bipedal locomotion—from robots to humans. AIP J. Nonlinear Sci. 2009, 19, 026101.
[CrossRef]
Sensors 2022, 22, 4440 26 of 31

35. Macleod, C.A.; Meng, L.; Conwayand, B.A.; Porr, B. Reflex control of robotic gait using human walking data. PLoS ONE 2014,
9, e109959. [CrossRef]
36. Haberland, M.; Kim, S. On extracting design principles from biology: II. Case study—The effect of knee direction on bipedal
robot running efficiency. Bioinspir. Biomim. 2015, 10, 016011. [CrossRef] [PubMed]
37. Renjewski, D.; Seyfarth, A. Robots in human biomechanics a study on ankle push-off in walking. Bioinspir. Biomim. 2012, 7, 036005.
[CrossRef] [PubMed]
38. Goswami, A. Postural stability of biped robots and the foot-rotation indicator (FRI) point. Int. J. Robot. Res. 1999, 18, 523–533.
[CrossRef]
39. McGeer, T. Passive walking with knees. In Proceedings of the IEEE International Conference on Robotics and Automation,
Cincinnati, OH, USA, 13–18 May 1990; Volume 3, pp. 1640–1645. [CrossRef]
40. Fujiki, S.; Aoi, S.; Funato, T.; Tomita, N.; Senda, K.; Tsuchiya, K. Adaptation mechanism of interlimb coordination in human
split-belt treadmill walking through learning of foot contact timing: A robotics study. J. R. Soc. Interface 2015, 12, 0542. [CrossRef]
[PubMed]
41. Reisman, D.; Wityk, R.; Silver, K.; Bastian, A.J. Locomotor adaptation on a split-belt treadmill can improve walking symmetry
post-stroke. Brain 2007, 130, 1861–1872. [CrossRef] [PubMed]
42. Haddad, J.M.; van Emmerik, N.; Whittlesey, R.E.; Hamill, J. Adaptations in interlimb and intralimb coordination to asymmetrical
loading in human walking. Gait Posture 2006, 23, 429–434. [CrossRef] [PubMed]
43. Reisman, D.; Block, H.J.; Bastian, A.J. Interlimb coordination during locomotion: What can be adapted and stored? J. Neurophysiol.
2005, 94, 2403–2415. [CrossRef]
44. Torres-Oviedo, G.; Vasudevan, E.; Malone, L.; Bastian, A.J. Locomotor adaptation. In Progress in Brain Research; Elsevier:
Amsterdam, The Netherlands, 2011; Volume 191, pp. 65–74. [CrossRef]
45. Reisman, D.; Bastian, A.J.; Morton, M. Neurophysiologic and rehabilitation insights from the split-belt and other locomotor
adaptation paradigms. Phys. Ther. 2010, 90, 187–195. [CrossRef]
46. Lim, H.; Takanishi, A. Compensatory motion control for a biped walking robot. Robotica 2005, 23, 1–11. [CrossRef]
47. Al-Yahya, W.; Mahmoud, D.; Meester, E.; Esser, P.; Dawes, H. Neural substrates of cognitive motor interference during walking;
peripheral and central mechanisms. Front. Hum. Neurosci. 2019, 12, 536. [CrossRef]
48. Seireg, A.; Arvikar, R. A mathematical model for evaluation of forces in lower extremeties of the musculo-skeletal system.
J. Biomech. 1973, 6, 313–326. [CrossRef]
49. Townsend, M.A.; Seireg, A.A. Effect of model complexity and gait criteria on the synthesis of bipedal locomotion. IEEE Trans.
Biomed. Eng. 1973, 6, 433–444. [CrossRef]
50. Migliore, A. The Role of Passive Joint Stiffness and Active Knee Control in Robotic Leg Swinging: Applications to Dynamic
Walking. Ph.D. Thesis, Georgia University, Athens, GA, USA, 2008.
51. Sangwan, V.; Agrawal, K. Differentially flat design of bipeds ensuring limit cycles. IEEE Trans. Mechatron. 2009, 14, 647–657.
[CrossRef]
52. Chevallereau, C.; Bessonnet, G.; Abba, G.; Aoustin, Y. Bipedal Robots: Modeling, Design and Walking Synthesis; John & Wiley & Sons:
Hoboken, NJ, USA, 2013; p. 338. ISBN 978-1-118-62297-1.
53. Miyazaki, F.; Arimoto, S. A Control theoretic study on dynamical biped locomotion. ASME J. Dyn. Sys. Meas. 1980, 102, 233–239.
[CrossRef]
54. Dekker, M. Zero-moment point method for stable biped walking. Eindh. Univ. Technol. 2009, 62, 0610544.
55. Whittle, M.W. Gait Analysis: An Introduction; Butterworth-Heinemann: Oxford, UK, 2014; p. 224.
56. van Zutven, P.; Kostić, D.; Nijmeijer, H. On the stability of bipedal walking. In Lecutre Notes in Computer Sciences; Springer:
Berlin/Heidelberg, Germany, 2010; pp. 521–532. [CrossRef]
57. Zielińska, T. Biological inspiration used for robots motion synthesis. J. Physiol. Paris 2009, 103, 133–140. [CrossRef]
58. Nicholls, E. Bipedal Dynamic Walking in Robotics. Ph.D. Thesis, The University of Western Australia, Crawley, WA, Australia,
1998; 86p. Available online: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.590.6362&rep=rep1&type=pdf (ac-
cessed on 27 May 2022).
59. Pratt, J.E.; Tedrake, R. Velocity-based stability margins for fast bipedal walking. In Fast Motions in Biomechanics and Robotics;
Springer: Berlin/Heidelberg, Germany, 2006; pp. 299–324. [CrossRef]
60. Pratt, J.; Carff, J.; Drakunov, S.; Goswami, A. Capture point: A step toward humanoid push recovery. In Proceedings of the 6th
IEEE-RAS International Conference on Humanoid Robots, Genova, Italy, 4–6 December 2006; pp. 200–207. [CrossRef]
61. Wight, D.L.; Kubica, E.G.; Wang, D.W. Introduction of the foot placement estimator: A dynamic measure of balance for bipedal
robotics. J. Comput. Nonlinear Dyn. 2008, 3, 011009. [CrossRef]
62. Znegui, W.; Gritli, H.; Belghith, S. Analysis and control of the dynamic walking of the compass biped walker using poincaré
maps: Comparison between two design approaches. In Proceedings of the 2021 IEEE 2nd International Conference on Signal,
Control and Communication (SCC), Tunis, Tunisia, 20–22 December 2021; pp. 225–233. [CrossRef]
63. Shiriaev, A.S.; Freidovich, L.B. Transverse linearization for impulsive mechanical systems with one passive link. IEEE Trans.
Autom. Control. 2009, 54, 2882–2888. [CrossRef]
64. Iqbal, S.; Zang, X.; Zhu, Y.; Zhao, J. Bifurcations and chaos in passive dynamic walking: A review. Robot. Auton. Syst. 2014, 62,
889–909. [CrossRef]
Sensors 2022, 22, 4440 27 of 31

65. He, B.; Si, Y.; Wang, Z.; Zhou, Y. Hybrid CPG–FRI dynamic walking algorithm balancing agility and stability control of biped
robot. Auton. Robots. 2019, 43, 1855–1865. [CrossRef]
66. Shafei, H.R.; Sadeghnejad, S.; Bahrami, M.; Baltes, J. A comparative study and development of a passive robot with improved
stability. In RoboCup 2014: Robot World Cup XVIII. RoboCup 2014. Lecture Notes in Computer Science; Bianchi, R., Akin, H.,
Ramamoorthy, S., Sugiura, K., Eds.; Springer: Cham, Switzerland, 2015; Volume 8992. [CrossRef]
67. McGeer, T. Passive dynamic walking. Int. J. Robot. Res. 1990, 9, 62–82. [CrossRef]
68. McGeer, T. Principles of walking and running. Adv. Comp. Environ. Physiol. 1992, 11, 113–139.
69. Znegui, W.; Gritli, H.; Belghith, S. Stabilization of the passive walking dynamics of the compass-gait biped robot by developing
the analytical expression of the controlled Poincaré map. Nonlinear Dyn. 2020, 101, 1061–1091. [CrossRef]
70. Goswami, A.; Espiau, B.; Keramane, A. Limit cycles and their stability in a passive bipedal gait. In Proceedings of the IEEE
International Conference on Robotics and Automation, Minneapolis, MN, USA, 22–28 April 1996.
71. Garcia, M.; Chatterjee, A.; Ruina, A.; Coleman, M. The simplest walking model: Stability, complexity, and scaling. J. Biomech. Eng.
1998, 120, 281–288. [CrossRef] [PubMed]
72. Howell, G.W.; Baillieul, J. Simple controllable walking mechanisms which exhibit bifurcations. In Proceedings of the 37th IEEE
Conference on Decision and Control, Tampa, FL, USA, 18 December 1998.
73. Uchida, K.; Furuta, K. Constant torque walking. In Proceedings of the 2000 IEEE International Conference on Control Applications,
Anchorrage, AK, USA, 27 September 2000.
74. Osuka, K.; Fujitani, T.; Ono, T. Passive walking robot QUARTET. In Proceedings of the 1999 IEEE International Conference on
Control Applications, Kohala Coast, HI, USA, 22–27 August 1999.
75. Asano, F.; Zhi-Wei, L. On efficiency and optimality of asymmetric dynamic bipedal gait. In Proceedings of the IEEE International
Conference on Robotics and Automation, Kobe, Japan, 12–17 May 2009.
76. Asano, F. Efficiency analysis of 2-period dynamic bipdal gaits. In Proceedings of the IEEE/RSJ International Conference on
Intelligent Robots and Systems, St. Louis, MO, USA, 10–15 October 2009.
77. Kajita, S.; Kanehiro, F.; Kaneko, K.; Fujiwara, K.; Harada, K.; Yokoi, K.; Hirukawa, H. Biped walking pattern generation by using
preview control of zero-moment point. In Proceedings of the IEEE International Conference on Robotics and Automation, Taipei,
Taiwan, 14–19 September 2003; Volume 2, pp. 1620–1626. [CrossRef]
78. Yin, Y.; Hosoe, S. Mixed Logic dynamical modeling and on line optimal control of biped robot. In Proceedings of the 2006
IEEE/RSJ International Conference on Intelligent Robots and Systems, Beijing, China, 9–15 October 2007; pp. 315–328. [CrossRef]
79. Azevedo, C.; Poignet, P.; Espiau, B. On line optimal control for biped robots. IFAC Proc. Vol. 2002, 35, 199–204. [CrossRef]
80. Wieber, P.-B. Trajectory free linear model predictive control for stable walking in the presence of strong perturbations. In Proceed-
ings of the 2006 6th IEEE-RAS International Conference on Humanoid Robots, Genova, Italy, 4–6 December 2006; pp. 137–142.
[CrossRef]
81. Scianca, N.; De Simone, D.; Lanari, L.; Oriolo, G. MPC for humanoid gait generation: Stability and feasibility. IEEE Trans. Robot.
2020, 36, 1171–1188. [CrossRef]
82. Venkatesan, K.P.; Mahendrakar, P.R.; Mohan, R. Solving Inverse Kinematics using Geometric Analysis for Gait Generation in
Small-Sized Humanoid Robots. In Proceedings of the 2020 IEEE/SICE International Symposium on System Integration (SII),
Honolulu, HI, USA, 12–15 January 2020; pp. 384–389. [CrossRef]
83. Vundavilli, P.R.; Pratihar, D.K. Gait planning of biped robots using soft computing: An attempt to incorporate intelligence.
In Intelligent Autonomous Systems; Springer: Berlin/Heidelberg, Germany, 2010; pp. 57–85.
84. Zheng, Y.-F. A neural gait synthesizer for autonomous biped robots. In Proceedings of the IEEE International Workshop on
Intelligent Robots and Systems, Towards a New Frontier of Applications, Ibaraki, Japan, 3–6 July 1990; pp. 601–608. [CrossRef]
85. Sugihara, T. Mobility Enhancement Control of Humanoid Robot Based on Reaction Force Manipulation via Whole Body Motion.
Ph.D. Thesis, University of Tokyo, Tokyo, Japan, 2004; 163p.
86. Yüksel, B. Towards the Enhancement of Biped Locomotion and Control Techniques. Ph.D. Thesis, The Graduate School of Natural
and Applied Sciences of Middle East Technical University, Ankara, Turkey, 2008; 328p.
87. Thulborn, R.A. Speeds and gaits of dinosaurs. Palaeogeogr. Palaeoclimatol. Palaeoecol. 1982, 38, 227–256. [CrossRef]
88. Alexander, R.M. The gaits of bipedal and quadrupedal animals. Int. J. Robot. Reearch 1984, 3, 49–59. [CrossRef]
89. Alexander, R.M. Bipedal animals, and their differences from humans. J. Anat. 2004, 204, 321–330. [CrossRef]
90. Abourachid, A.; Hugel, V. The natural bipeds, birds and humans: An inspiration for bipedal robots. In Proceedings of the 5th
International Conference, Living Machines, Edinburgh, UK, 19–22 July 2016; pp. 3–15. [CrossRef]
91. Daley, M.A. Understanding the agility of running birds: Sensorimotor and mechanical factors in avian bipedal locomotion. Integr.
Comp. Biol. 2018, 58, 884–893. [CrossRef]
92. Daley, M.A.; Birn-Jeffery, A. Scaling of avian bipedal locomotion reveals independent effects of body mass and leg posture on gait.
J. Exp. Biol. 2018, 221 Pt 10, jeb152538. [CrossRef] [PubMed]
93. Lim, H.-O.; Takanishi, A. Biped walking robots created at Waseda University: WL and WABIAN family. Philos. Trans. R. Soc.
Math. Phys. Eng. Sci. 2006, 365, 49–64. [CrossRef] [PubMed]
94. HRP-2 Robot. Available online: https://robots.ieee.org/robots/hrp2/ (accessed on 15 January 2021).
95. iCub Robot. Available online: https://robots.ieee.org/robots/icub/ (accessed on 15 January 2021).
Sensors 2022, 22, 4440 28 of 31

96. Gouaillier, D.; Hugel, V.; Blaz, P.; Kilner, C.; Monceaux, J.; Lafourcade, P.; Marnier, B.; Serre, J.; Maisonnier, B. Mechatronic
design of NAO humanoid. In Proceedings of the 2009 IEEE InternationalConference on Robotics and Automation, Kobe, Japan,
12–17 May 2009; pp. 769–774. [CrossRef]
97. HRP-4C. Available online: https://robots.ieee.org/robots/hrp4c/ (accessed on 15 January 2021).
98. HRP-4. Available online: https://robots.ieee.org/robots/hrp4/ (accessed on 15 January 2021).
99. Nelson, G.M.; Saunders, A.; Neville, N.; Swilling, B.J.; Bondaryk, J.E.; Billings, D.; Lee, C.; Playter, R.; Raibert, M.H. Petman:
A humanoid robot for testing chemical protective clothing. J. RoboSoc. Jpn. 2012, 30, 372–377. [CrossRef]
100. REEM-C. Available online: https://robots.ieee.org/robots/reemc/ (accessed on 15 January 2021).
101. Debiut of Atlas Robot. DARPA. Available online: https://www.darpa.mil/about-us/timeline/debut-atlas-robot (accessed on
19 August 2020).
102. Guizzo, E. By leaps and bounds: An exclusive look at how boston dynamics is redefining robot agility. IEEE Spectr. 2019, 56,
34–39. [CrossRef]
103. Boston Dynamics’ Atlas Robot Now Does Gymnastics, Too. Available online: https://www.youtube.com/watch?v=kq6mJOktIvM
(accessed on 8 February 2020).
104. Robonaut Leg in Motion. 2014. Available online: https://www.youtube.com/watch?v=12iayXdNTdM (accessed on 30 January 2021).
105. Englsberger, J.; Werner, A.; Ott, C.; Henze, B.; Roa, M.A.; Garofalo, G.; Burger, R.; Beyer, A.; Eiberger, O.; Schmid, K.; et al.
Overview of the torque-controlled humanoid robot TORO. In Proceedings of the 2014 IEEE-RAS International Conference on
Humanoid Robots, Madrid, Spain, 18–20 November 2014; IEEE: Piscataway, NJ, USA; pp. 916–923. [CrossRef]
106. Robotic Ostrich to the Rescue. Available online: https://www.audubon.org/news/robotic-ostrich-rescue (accessed on
30 August 2021).
107. WALK-MAN Team Built Brand New, Highly Custom Robot for DRC Finals—(2015). Available online: https://spectrum.ieee.org/
walkman-humanoid-robot-iit (accessed on 30 January 2021).
108. Stentz, A.; Herman, H.; Kelly, A.; Meyhoferet, E.; Haynes, G.C.; Stager, D.; Zajac, B.; Bagnell, J.A.; Brindza, J.; Dellin, C. CHIMP,
the CMU highly intelligent mobile platform. J. Field Robot. 2015, 32, 209–228. [CrossRef]
109. Kim, S.; Kim, M.; Lee, J.; Hwang, S.; Chae, J.; Park, B.; Cho, H.; Sim, J.; Jung, J.; Lee, H.; et al. Approach of team SNU to the
DARPA robotics challenge finals. In Proceedings of the 2015 IEEE-RAS 15th International Conference on Humanoid Robots
(Humanoids), Seoul, Korea, 3–5 November 2015; pp. 777–784. [CrossRef]
110. THORMANG Robot. Available online: https://youtu.be/B9myJuyoibM (accessed on 30 July 2021).
111. Radford, N.A.; Strawser, P.; Hambuchen, K.; Mehling, J.S.; Verdeyen, W.K.; Donnan, A.S.; Holley, J.; Sanchez, J.; Nguyen, V.;
Bridgwater, L.; et al. Valkyrie: NASA’s first bipedal humanoid robot. J. Field Robot. 2015, 32, 397–419. [CrossRef]
112. DRC-HUBO+. Available online: https://robots.ieee.org/robots/drchubo/ (accessed on 15 January 2021).
113. DURUS: SRI’s Ultra-Efficient Walking Humanoid Robot. Available online: https://spectrum.ieee.org/durus-sri-ultra-efficient-
humanoid-robot (accessed on 15 January 2022).
114. SRI’S DURUS. Available online: https://www.youtube.com/watch?v=HyqT9Bdamt8 (accessed on 15 January 2022).
115. Wu, L.; Larkin, M.; Potnuru, A.; Tadesse, Y. HBS-1: A Modular Child-Size 3D Printed Humanoid. Robotics 2016, 5, 1. [CrossRef]
116. Asano, Y.; Kozuki, T.; Ookubo, S.; Kawamura, M.; Nakashima, S.; Katayama, T.; Yanokura, I.; Hirose, T.; Kawaharazuka,
K.; Makino, S.; et al. Human mimetic musculoskeletal humanoid Kengoro toward real world physically interactive actions.
In Proceedings of the 2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids), Cancun, Mexico,
15–17 November 2016; pp. 876–883. [CrossRef]
117. Kaminaga, H.; Ko, T.; Masumura, R.; Komagata, M.; Sato, S.; Yorita, S.; Nakamura, Y. Mechanism and control of whole-body
electro-hydrostatic actuator driven humanoid robot hydra. In International Symposium on Experimental Robotics; Springer: Cham,
Switzerland, 2016; pp. 656–665. [CrossRef]
118. Cassie. Available online: https://robots.ieee.org/robots/cassie/ (accessed on 15 January 2021).
119. Ficht, G.; Allgeuer, P.; Farazi, H.; Behnke, S. NimbRo-OP2: Grown-up 3D printed open humanoid platform for research.
In Proceedings of the 2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids), Birmingham, UK,
15–17 November 2017; pp. 669–675. [CrossRef]
120. TALOS. Available online: https://robots.ieee.org/robots/talos (accessed on 15 August 2021).
121. HRP-5P. Available online: https://robots.ieee.org/robots/hrp5p/ (accessed on 15 January 2021).
122. Digit. Available online: https://robots.ieee.org/robots/digit/ (accessed on 15 January 2021).
123. Hobart, C.G.; Mazumdar, A.; Spencer, S.J.; Quigley, M.; Smith, J.P.; Bertrand, S.; Pratt, J.; Kuehl, M.; Buerger, S.P. Achieving
Versatile Energy Efficiency With the WANDERER Biped Robot. IEEE Trans. Robot. 2020, 36, 959–966. [CrossRef]
124. WANDERER. Available online: https://www.youtube.com/watch?v=tIm0muIOc5E (accessed on 15 January 2022).
125. Ha, I.; Tamura, Y.; Asama, H.; Han., H.; Hong, D.W. Development of open humanoid platform DARwIn-OP. In Proceedings of the
SICE Annual Conference (SICE), Tokyo, Japan, 13–18 September 2011; pp. 2178–2181.
126. Lapeyre, M.; Rouanet, P.; Grizou, J.; N’Guyen, S.; Depraetre, F.; Falher, A.L.; Oudeyer, P. Poppy Project: Open-Source Fabrication of
3D Printed Humanoid Robot for Science, Education and Art; Digital Intelligence: Nantes, France, 2014; 6p.
127. Lim, C.; Yeap, G.H. The locomotion of bipedal walking robot with six degree of freedom. Procedia Eng. 2012, 41, 8–14. [CrossRef]
128. RQ HUNO. Available online: https://www.robobuilder.net/rq-huno (accessed on 19 August 2020).
Sensors 2022, 22, 4440 29 of 31

129. Red Dragon V3. Available online: https://www.robotshop.com/community/forum/t/red-dragon-v3/5340 (accessed on

1 September 2020).
130. w00dBob—A Biped Wooden Robot Made with Arduino Nano. Available online: https://blog.adafruit.com/2014/03/29/w0
0dbob-a-biped-wooden-robot-made-with-arduino-nano-arduinod14/ (accessed on 1 September 2020).
131. Mikolajczyk, T.; Borboni, A.; Kong, X.; Malinowski, T.; Olaru, A. 3D printed biped walking robot. Appl. Mech. Mater. 2015, 772,
477–481. [CrossRef]
132. Mikolajczyk, T.; Fas, T.; Malinowski, T.; Romanowski, Ł. Prototype model of walking robot. Appl. Mech. Mater. 2014, 613, 21–28.
[CrossRef]
133. Mikolajczyk, T.; Fas, T.; Malinowski, T.; Romanowski, Ł. New solution of Walking Robot. Appl. Mech. Mater. 2014, 555, 232–238.
[CrossRef]
134. 4DOF Robot. Available online: https://youtu.be/yhWYoV3waG0 (accessed on 16 December 2021).
135. Wang, K.; Shah, A.; Kormushev, P. SLIDER: A Novel Bipedal Walking Robot without Knees. In Proceedings of the 19th
International Conference towards Autonomous Robotic Systems (TAROS 2018), Bristol, UK, 25–27 July 2018.
136. Slider Walking Robott Imperial College Robot Intelligence Lab. Available online: https://www.youtube.com/watch?v=
DeFMWLdCYng (accessed on 29 August 2020).
137. Kim, K.; Spieler, P.; Lupu, E.-S.; Ramezani, A.; Chung, S.-J. A bipedal walking robot that can fly, slackline, and skateboard.
Sci. Robot. 2021, 6, eabf8136. [CrossRef] [PubMed]
138. Meet Leo, a Robot with Multiple Skills. Available online: https://www.youtube.com/watch?v=xI6IGkrLjho (accessed on
1 March 2022).
139. U-M First in Line For New Bird-Inspired Walking Robot. Available online: https://news.umich.edu/u-m-first-in-line-for-new-
bird-inspired-walking-robot/ (accessed on 1 March 2022).
140. Van der Linde, R. Towards applicable ballistic walking. Technol. Health Care 1999, 7, 449–453. [CrossRef] [PubMed]
141. Moon, J.; Bae, J. Gait optimization and energetics of ballistic walking for an underactuated biped with knees. Nonlinear Dyn. 2016,
85, 1533–1546. [CrossRef]
142. Darici, O.; Temeltas, H.; Kuo, A.D. Optimal regulation of bipedal walking speed despite an unexpected bump in the road.
PLoS ONE 2018, 13, e0204205. [CrossRef]
143. Rebula, J.R.; Schaal, S.; Finley, J.; Righetti, L. A Robustness analysis of inverse optimal control of bipedal walking. IEEE Robot.
Autom. Lett. 2019, 4, 4531–4538. [CrossRef]
144. de Pina, A.; Filho, C.; Dutra, M.; Raptopoulos, L. Modeling of a bipedal robot using mutually coupled Rayleigh oscillators. Biol.
Cybern. 2005, 92, 1–7. [CrossRef]
145. Luo, Q.; Pan, C.; Wu, G. Central pattern generator–based coupling control method for synchronously controlling the two-degrees-
of-freedom robot. Sci. Prog. 2020, 103. [CrossRef]
146. Chen, C.-P.; Chen, J.-Y.; Huang, C.-K.; Lu, J.-C.; Lin, P.-C. Sensor data fusion for body state estimation in a bipedal robot and its
feedback control application for stable walking. Sensors 2015, 15, 4925–4946. [CrossRef]
147. Klein, J.; Lewis, M.A. A physical model of sensorimotor interactions during locomotion. J. Neural Eng. 2012, 9, 046011. [CrossRef]
[PubMed]
148. Forner-Cordero, M.; Ackermann, A.; de Lima Freitas, M. A method to simulate motor control strategies to recover from
perturbations: Application to a stumble recovery during gait. In Proceedings of the International Conference of the IEEE
Engineering in Medicine and Biology Society, Boston, MA, USA, 30 August–3 September 2011; pp. 7829–7832. [CrossRef]
149. Zajac, F.E.; Neptune, R.R.; Kautz, A. Biomechanics and muscle coordination of human walking: Part II: Lessons from dynamical
simulations and clinical implications. Gait Posture 2003, 17, 1–17. [CrossRef]
150. Kakou, H.; Shitama, H.; Kimura, Y.; Nakamoto, Y.; Furuta, N.; Honda, K.; Wada, F.; Hachisuka, H. Walking assist robot and its
clinical application. J. UOEH 2009, 31, 207–218. [CrossRef] [PubMed]
151. Cifuentes, C.A.; Frizera, A. Human-Robot Interaction Strategies for Walker-Assisted Locomotion; Springer: Cham, Switzerland, 2016;
Volume 115. [CrossRef]
152. Takahashi, R.; Hosoda, K. Reflex-Based Walking Controller for Real Bipedal Robot—From Phase-Based to Reflex-Based. 2019.
Available online: https://infoscience.epfl.ch/record/272162/files/Extended%20abstract.pdf (accessed on 27 May 2022).
153. Geng, T.; Porr, B.; Wörgötter, F. A reflexive neural network for dynamic biped walking control. Neural Comput. 2006, 18, 1156–1196.
[CrossRef] [PubMed]
154. Pfeiffer, F. Technological aspects of walking. In Walking: Biological and Technological Aspects; Pfeiffer, F., Zielinska, T., Eds.; Springer
Verlag Wien GmbH: Wien, Austria, 2004. [CrossRef]
155. Sooyong, L.; Oh Paul, Y. Sensor Information Analysis for a Humanoid Robot. Int. J. Control Autom. Syst. 2015, 13, 175–181.
[CrossRef]
156. Ishida, T.; Kuroki, Y. Sensor system of a small biped entertainment robot. Adv. Robot. 2004, 18, 1039–1052. [CrossRef]
157. Hashimoto, K. A review on vision-based control of robot manipulators. Adv. Robot. 2003, 17, 969–991. [CrossRef]
158. Rojas-Quintero, J.A.; Rodríguez-Liñán, M.C. A literature review of sensor heads for humanoid robots. Robot. Auton. Syst. 2021,
143, 103834. [CrossRef]
159. Playter, R. The PETMAN and atlas robots at boston dynamics. In Humanoid Robotics: A Reference; Goswami, A., Vadakkepat, P.,
Eds.; Springer: Dordrecht, The Netherlands, 2019; pp. 169–186. [CrossRef]
Sensors 2022, 22, 4440 30 of 31

160. Fallón, M.F.; Antone, M.; Roy, N.; Teller, S. Drift-free humanoid state estimation fusing kinematic, inertial and LIDAR sensing.
In Proceedings of the 2014 IEEE-RAS International Conference on Humanoid Robots, Madrid, Spain, 18–20 November 2014;
pp. 112–119. [CrossRef]
161. Fallon, M.F.; Marion, P.; Deits, R.; Whelan, T.; Antone, M.; McDonald, J.; Tedrake, R. Continuous humanoid locomotion over
uneven terrain using stereo fusion. In Proceedings of the 2015 IEEE-RAS 15th International Conference on Humanoid Robots
(Humanoids), Seoul, Korea, 3–5 November 2015; pp. 881–888. [CrossRef]
162. Piperakis, S.; Kanoulas, D.; Tsagarakis, N.G.; Trahanias, P. Outlier-robust state estimation for humanoid robots. In Proceedings
of the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Macau, China, 3–8 November 2019;
pp. 706–713. [CrossRef]
163. Camurri, M.; Ramezani, M.; Nobili, S.; Fallon, M.A. Multi-Sensor State Estimator for Legged Robots in Real-World Scenarios.
Front. Robot. AI 2020, 7, 68. [CrossRef]
164. Yang, G.; Bellingham, J.; Dupont, P.; Fischer, P.; Floridi, L.; Full, R.; Jacobstein, N.; Kumar, V.; McNutt, M.; Merrifield, R.; et al. The
grand challenges of science robotics. Sci. Robot. 2018, 3, eaar7650. [CrossRef]
165. May, G.J.; Davidson, A.; Monahov, B. Lead batteries for utility energy storage: A review. J. Energy Storage 2018, 15, 145–157.
[CrossRef]
166. Great Power Group. Square Lithium-Ion Cell. Available online: http://www.greatpower.net/cplb/info_159.aspx?itemid=292&
cid=25 (accessed on 31 December 2019).
167. Lithium-Ion Battery. Available online: https://www.cei.washington.edu/education/science-of-solar/battery-technology/
(accessed on 26 May 2022).
168. Ultra-Light Battery Pack for Walking Robot. Available online: https://www.cie-solutions.com/robotics-battery-pack (accessed
on 27 May 2022).
169. Budde-Meiwes, H.; Drillkens, J.; Lunz, B.; Muennix, J.; Rothgang, S.; Kowal, J.; Sauer, D.U. A review of current automotive battery
technology and future prospects. Proc. Inst. Mech. Eng. Part D J. Automob. Eng. 2013, 227, 761–776. [CrossRef]
170. Jones, J.L.; Bruce, A.S.; Anita, M.F. Mobile Robots: Inspiration to Implementation; CRC Press: Boca Raton, FL, USA, 1998.
171. Hydrogen Cyllindee General Specification. Available online: https://www.qtww.com/wp-content/uploads/2019/01/H2-Tank-
Specifications-Sept-2021-All-Tanks.pdf (accessed on 27 May 2022).
172. This Star Wars-Inspired Walking Robot Weighs Two Tonnesand Travels at Less Than 1 Mph. Available online: https:
//www.guinnessworldrecords.com/news/2018/9/video-this-star-wars-inspired-walking-robot-weighs-two-tonnes-and-
travels-at-les-540955 (accessed on 27 May 2022).
173. Amiryar, M.E.; Keith, R.P. A review of flywheel energy storage system technologies and their applications. Appl. Sci. 2017, 7, 286.
[CrossRef]
174. BigDog—The Most Advanced Rough-Terrain Robot on Earth. Boston Dynamics. Available online: https://web.archive.org/
web/20170518204101/http:/www.bostondynamics.com/robot_bigdog.html (accessed on 27 May 2022).
175. Thangavelautham, J.; Gallardo, D.; Strawser, D.; Dubowsky, S. Hybrid fuel cell power for long duration robot missions in field
environments. In Field Robotics; World Scientific: Singapore, 2011; pp. 471–478. [CrossRef]
176. RoBeetle: A Micro Robot Powered by Liquid Fuel. Available online: https://www.youtube.com/watch?v=vd6tN19wilQ
(accessed on 26 May 2022).
177. Can Photovoltaics Power Humanoid Robots? Available online: https://www.azorobotics.com/Article.aspx?ArticleID=328
(accessed on 27 May 2022).
178. Robotics—Power Source for Robots, 2013. Available online: https://www.azorobotics.com/Article.aspx?ArticleID=139 (accessed
on 22 May 2022).
179. Meet Tertill. Available online: https://www.youtube.com/watch?v=VwTWhMbnq9g (accessed on 27 May 2022).
180. Vitirover. Available online: https://www.pv-magazine.com/2022/02/24/solar-powered-robotic-mower-for-vineyards/ (ac-
cessed on 27 May 2022).
181. Solar Walking Robot. Available online: https://www.youtube.com/watch?v=JSGqPvRPWEY (accessed on 26 May 2022).
182. Future Batteries, Coming Soon: Charge in Seconds, Last Months and Power Over the Air, Mart 2021. Available online:
https://www.pocket-lint.com/gadgets/news/130380-future-batteries-coming-soon-charge-in-seconds-last-months-and-
power-over-the-air (accessed on 26 May 2022).
183. Tucker, V.A. The energetic cost of moving about: Walking and running are extremely inefficient forms of locomotion. Much
greater efficiency is achieved by birds, fish—And bicyclists. Am. Sci. 1975, 63, 413–419.
184. Kajita, S.; Espiau, B. Legged robots. In Springer Handbook of Robotics; Siciliano, B., Khatib, O., Eds.; Springer: Berlin/Heidelberg,
Germany, 2008.
185. Sakagami, Y.; Watanabe, R.; Aoyama, C.; Matsunaga, S.; Higaki, N.; Fujimura, K. The intelligent ASIMO: System overview
and integration. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IEEE), Lausanne,
Switzerland, 30 September–4 October 2002; Volume 3, pp. 2478–2483. [CrossRef]
186. Kashiri, N.; Abate, A.; Abram, S.J.; Albu-Schaffer, A.; Clary, P.J.; Daley, M.; Faraji, S.; Furnemont, R.; Garabini, M.; Geyer, H.; et al.
An overview on principles for energy efficient robot locomotion. Front. Robot. AI 2018, 5, 129. [CrossRef] [PubMed]
Sensors 2022, 22, 4440 31 of 31

187. Reher, J.; Cousineau, E.A.; Hereid, A.; Hubicki, C.M.; Ames, A.D. Realizing dynamic and efficient bipedal locomotion on
the humanoid robot DURUS. In Proceedings of the 2016 IEEE International Conference on Robotics and Automation (ICRA),
Stockholm, Sweden, 16–21 May 2016; pp. 1794–1801.
188. Meet Cassie—The Highly Energy Efficient Walking Machine. Available online: https://www.borntoengineer.com/meet-cassie-
highly-energy-efficient-walking-machine (accessed on 15 February 2022).
189. Cassie Robot Runs Turf 5k. Available online: https://www.youtube.com/watch?v=FSaSjd_HOaI (accessed on 30 August 2021).
190. Li, W.; Szczecinski, N.S.; Quinn, R.D. A neural network with central pattern generators entrained by sensory feedback controls
walking of a bipedal model. Bioinspir. Biomim. 2017, 12, 065002. [CrossRef] [PubMed]
191. Wu, Q.; Yang, X.; Wu, Y.; Zhou, Z.; Wang, J.; Zhang, B.; Luo, Y.; Chepinskiy, S.A.; Zhilenkov, A.A. A novel underwater bipedal
walking soft robot bio-inspired by the coconut octopus. Bioinspir. Biomim. 2021, 16, 046007. [CrossRef]
192. Leal, P.B.C.; Cabral-Seanez, M.; Baliga, V.B.; Altshuler, D.L.; Hartl, D.J. Phase transformation-driven artificial muscle mimics the
multifunctionality of avian wing muscle. J. R. Soc. Interface 2021, 18, 1042. [CrossRef]
193. Voigt, M.; Hansen, E.A. The puzzle of the walk-to-run transition in humans. Gait Posture 2021, 86, 319–326. [CrossRef]
194. Samani, F.; Ceccarelli, M. Prototype design and testing of TORVEastro, cable-driven astronaut robot. In Proceedings of theIn-
ternational Conference on Robotics in Alpe-Adria Danube Region, Kaiserslautern, Germany, 19 June 2020; Springer: Cham,
Switzerland, 2020; pp. 448–455.
195. Smyrli, A.; Bertos, G.A.; Papadopoulos, E. A generalized model for compliant passive bipedal walking: Sensitivity analysis and
implications on bionic leg design. J. Biomech. Eng. 2021, 143, 101008. [CrossRef]