Received November 19, 2018, accepted December 18, 2018, date of publication December 24, 2018,

date of current version January 16, 2019.

Digital Object Identifier 10.1109/ACCESS.2018.2889245

Development of Modular Cable-Driven

Parallel Robotic Systems
SEN QIAN, BIN ZI , DAOMING WANG, AND YUAN LI
School of Mechanical Engineering, Hefei University of Technology, Hefei 230009, China
Corresponding author: Bin Zi ([email protected])
This work was supported by the National Natural Science Foundation of China under Grant 51575150 and Grant 51605126.

ABSTRACT Task-oriented modular design and safety monitoring are demonstrated to be outstanding
challenges for robots in practical application. In order to obtain better flexibility and better obstacles
avoidance capability, design and analysis of a modular cable-driven parallel robot (MCDPR) are presented
in this paper, which can be reconfigured to several configurations in engineering applications by changing
the number of the mobile modules and the connection mode of the end effector. With regard to the motion
control and safety monitoring system, the design and implementation of hardware and software are presented.
An experimental prototype for the MCDPR is developed, followed by a brief illustration of the connection
mode between the end effector and cables, and the installation of monitoring nodes. The performance of
the MCDPR is discussed experimentally, including real-time pose monitoring and obstacle avoidance. The
results verify the feasibility and efficiency of kinematics model and obstacle avoidance method, and the
results also indicate that the designed motion control and safety monitoring system can realize trajectory
tracking control, real-time safety monitoring, and obstacle avoidance for the MCDPR.

INDEX TERMS Cable-driven parallel robots, design and implementation, safety monitoring, obstacle
avoidance.

I. INTRODUCTION requirements in terms of multifunction, high precision and

In the last decade, cable-driven parallel robots (CDPRs) have reliability, payload-to-weight ratio for the CDPRs, it is nec-
attracted a considerable amount of interest owing to a wide essary to explore the modular design of the task-oriented
area of applications and a lot of advantages, including large compound-driven CDPRs and safety monitoring system of
reachable workspace, high acceleration capability and eco- complex electromechanical system. Modular robots have
nomical construction [1]–[3]. In CDPR, flexible cables are received more and more attention to meet the requirement of
applied in place of rigid-link legs to support and manipulate flexibility in the manufacturing industry [18]–[21].
the end-actuator [4]. Due to these advantages, the CDPRs CDPRs offer many advantages considering the flexibil-
are eligible for applications like video capture system [5], ity of cables. However, cables can only actuate unilater-
large spherical radio telescope [6], large scale 3D printer [7], ally through tension, which limits the moment resisting
removing graffiti from highway signs [8], cooperative crane and exerting capabilities. Research in the topic of modular
system [9], periodic maintenance of suspension bridge [10], CDPRs (MCDPR) is highly motivated, which can enhance
humanoid robotic arm and musculoskeletal system [11], [12], the orientation workspaces of CDPRs and overcome the
and robot rehabilitation [13], [14]. All these CDPRs have drawback of CDPRs, especially the collisions between the
made valuable contributions to mechanical behavior and cables and the surrounding Environment [22]–[24]. Modular
mechatronics technology of the robotics. However, according design enables modular CDPRs to change their module and
to specific applications, most of them are concentrated on a size to meet specific operational demands and possess broad
single operational task of the CDPRs [15]. application prospects. For instance, multiple mobile cranes,
More recently, the realm of the CDPRs is beginning as shown in Fig. 1, can be used for material transportation,
to manifest its advantages, and high performance opera- weight elevation, steel erection and roof installation, etc. The
tion with high payload and great flexibility is an important cooperative lifting system consists of three and even more
task in modern engineering applications [6]–[9], [16], [17]. cranes can be seen as the possible applications of MCDPR by
Thus, under this situation, to meet the performance changing the number of the mobile cranes and the connection

2169-3536
2018 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 7, 2019 Personal use is also permitted, but republication/redistribution requires IEEE permission. 5541
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 S. Qian et al.: Development of MCDPR Systems

experimental studies. This paper is organized as follows.

Section II introduces the issues involved in the system
description of the MCDPR. The hardware and software
implementation of the motion control and safety monitoring
system is provided in Section III. Section IV describes the
development of the experimental testbed. The performances
of the MCDPR are evaluated and discussed in Section V.
Finally, conclusions are given in Section VI.

II. SYSTEM DESCRIPTION

Referring to Fig. 2, the MCDPR can be reconfigured to
several configurations to imitate different types of RCDPR
in engineering applications by changing the number of the
mobile modules and the connection mode of end-effector.
With four typical configurations shown in Fig. 2, the MCDPR
FIGURE 1. Applications of multiple mobile cranes.
are mainly consisted by a circular orbit, six groups of CDPRs,
six flexible cables and an end-effector. More specifically, six
mode of payload. In order to obtain better flexibility and mobile modules are equipped on a circular orbit. They can
obstacles avoidance capability, design and analysis of the rotate along a circular trajectory with respect to the center line
MCDPR are presented in this paper, which can be recon- of the orbit through gear transmission driven by a servomotor.
figured to several configurations of CDPRs in engineering A cable drum and a hydraulic cylinder with a fix pulley at the
applications by changing the number of the mobile modules top are both installed on each individual mobile module. The
and the connection mode of end-effector. After addressing the pulley component is freely rotatable about the vertical axis of
design issue, understanding and evaluating the mechanical the cylinder. For each cable, one end is rolled in a drum, and
behavior of the MCDPR is important for heterogeneous tasks the other end is directed to the end-effector through a pulley.
in different environments. Kinematics plays an important role The end-effector, employed as a controlled object, is actuated
in manipulator behavior. Kinematic analysis has been rela- by the cables to manipulate all DOFs.
tively presented in the work on manipulator robots [25], [26]. Obviously in Fig. 2, configuration I and III illustrates three-
Safety of the CDPRs is the important consideration before cable and four-cable point-mass CDPRs which is under-
it can be materialized. A number of studies investigate the constrained. Configurations II and IV are respectively the
design and implementation of system monitoring, safety three-cable and four-cable underconstrained CDPRs. Config-
strategy for human-robot collaboration system, and reliability uration V shows underconstrained CDPRs with six DOFs.
and safety of robot systems [27]–[30]. Consequently, it is Configuration VI shows underconstrained CDPRs with three
critical that only dependable robot monitoring systems are translational degree of freedom actuated by three parallel
deployed for human-robot collaborative tasks. Safety is the cable groups and a rotating degree of freedom around vertical
unified criterion for future technical challenges in the design direction actuated by mobile modules. Moreover, the vertical
and control of human-robot systems. However, mechanical stretch of the hydraulic cylinder and the rotation of the mobile
structures and physical characteristics of the CDPRs can module along the circular orbit can also alter configuration
hardly meet safe requirements absolutely and carry a high parameters of the MCDPR to improve flexibility and save
risk of injuries to operators. Additionally, in consideration costs for optimization [19].
of unpredictable obstacles within the operation environment,
obstacle avoidance proves to be a pressing problem that must III. DESIGN OF MOTION CONTROL AND
be solved for safety consideration [16]. SAFETY MONITORING SYSTEM
The primary motivation of this study is to investigate the A. MOTION CONTROL SYSTEM
design, safety monitoring, and obstacle avoidance of the Fig. 3 shows the hardware of the motion control system,
MCDPR in order to meet the performance requirements in which is composed of an industrial personal computer (IPC),
terms of multifunction, high reliability and payload-to-weight a motion controller, a pinboard and several servomotors and
ratio for the CDPRs. The contribution includes three aspects. drivers.
(1) A MCDPR is designed for better flexibility and better Basic mechanism and kinematic parameters should be con-
obstacles avoidance capability, which can be reconfigured figured prior to system startup. Specific values of parameters
to several configurations by changing the number of the are determined through measurements. After parameter con-
mobile modules and the connection mode of end-effector. figuration, the control system starts up and the host computer
(2) Mechatronic hardware and software for the experimental starts to upload the configuration data, eliminate the alarm
testbed are carried out with regard to the motion control information, set the motion mode of the shaft and switch
and safety monitoring system. (3) The effectiveness of the on the enabler for servomotors. Then, the motion module
safety monitoring and obstacle avoidance is verified through planning is performed based on the motion mode selection

5542 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

FIGURE 2. Several typical configurations of MCDPR: (a) configuration I; (b) configuration II;
(c) configuration III; (d) configuration IV; (e) configuration V; (f) configuration VI.

the safe requirements, which will cause severe injuries to

humans. Thus it is important to develop a monitoring system
to avoid collision and keep the security of operators through
sensors and signal processing techniques.

1) HARDWARE IMPLEMENTATION
Fig. 5 shows the hardware structure of the safety monitor-
ing system, which includes master control module, sensor
module, wireless transmission module and power source
FIGURE 3. Hardware configuration of the motion control system. module. The master control module, responsible for data
operation, storage and power consumption management,
after motor initialization. The motion mode typically con- is the core of the hardware system. A single chip micy-
tains the straight line mode and the spiral line mode. After oco (SCM) is selected as the microprocessor for master
motor position planning, the curves severed as the motor control module. Additionally, the sensors for obstacle mon-
input immediately plot in the software interface. Thereafter, itoring and end-effector pose monitoring, choose infrared
the motor stops upon the termination of the planning. Note obstacle avoidance (IOA) sensor and displacement sensor,
that a reset process is needed because the end-effector offsets respectively. For the wireless transmission module, a radio
its original pose after an experiment. frequency chip with a carrier frequency of 433 Hz is used for
The host computer interface for the motion control system the data forwarding from monitoring nodes to sink nodes. The
is designed as depicted in Fig. 4. It contains seven mod- power source module is in charge of the power supply for all
ules, which are respectively the basic mechanism parame- the other modules. A rechargeable Li-ion battery is a good
ter configuration module, the mode selection module, the candidate due to its compact size, long endurance and large
kinematic parameter configuration module, the mechanism capacity.
display module, the system startup and stop module, the plan- The monitoring circuit board includes the master control
ning curve display module, as well as the motor position module and the power module. For circuit diagram design of
planning module. the master control module, it contains the model selection of
microprocessor and the peripheral circuit design. The power
B. SAFETY MONITORING SYSTEM source module is used as a power supply for the circuit board
The MCDPR is designed for payload hoisting. As described and the wireless module. The working voltage of the circuit
in the introduction, the cooperative lifting system consists board and sensor module is 5V while that of the wireless
of three and even more cranes can be seen as the possible module is 3.3V.
applications of the MCDPR with different configurations,
which is a typical human-robot system. In the development 2) SOFTWARE IMPLEMENTATION
of human-robot systems, safety is the unified criterion for The software structure of the safety monitoring system is
future technical challenges. However, mechanical structures presented in Fig. 6. There are three steps from data acquisition
and physical characteristics of the CDPRs can hardly meet to data processing listed as:

VOLUME 7, 2019 5543

 S. Qian et al.: Development of MCDPR Systems

FIGURE 4. Host computer interface for motion control system.

FIGURE 6. Software structure of safety monitoring system.

FIGURE 5. Hardware structure of safety monitoring system.

range and angle of IOA are 80cm and 40◦ , respectively. The
Step 1 (Data Acquisition): The monitoring node is divided resolution of IOA is 50 µm. The output signal of the sensor
into the IOA monitoring (IOAM) node and the real-time pose has a direct nonlinear relation with the distance in the range
monitoring (RPM) node. They are separately applied for the from 0 to 10cm while an inverse nonlinear relation with the
monitoring of obstacle position, as well as the cable length distance from 10 to 80cm. The orbit radius is 60cm which
and the calculated real-time pose of the end-effector. falls within range from 10 to 80cm. However, the measuring
Step 2 (Data Sending): The wireless transmission method range of 0∼10cm can be regarded as a blind area because it
can overcome the shortcomings compared with the traditional locates in the orbit edge far away from the end-effector. Thus,
wire communication method, such as cockamamie wiring and the area can be considered as a relative safe area. Actually,
environmental sensitive. Firstly, the raw data is sent to the eight IOA sensors are used to achieve a full-scale monitoring
base station through wireless communication, and then it is of 360◦ since the effective measuring range of a single sensor
forwarded to the IPC through RS232 serial communication. is greater than 45◦ .
Step 3 (Data Processing and Display): The data packet The correspondence between output voltage and sensing
analysis is conducted by the host computer to gain required distance of IOA sensor is calibrated through several mea-
data for each module. Specifically, the data from IOA sensors surements, as shown in Fig. 7. The relationship curve in the
is converted to acquire the real-time position of obstacles. distance range from 10 to 80cm can be approximately treated
Therefore, the data is able to display directly on the host com- as six short straight lines. Accordingly, the relation between
puter. As for the real-time pose data, it is obtained through output voltage (Uout ) and sensing distance (d) follows the
the forward kinematic solution based on the feedback cable expression as
length. 
As is known to us, it is not allowed for any people or object  Line A: d = −8.33(Uout − 1.7)+15 Uout ≥ 1.7
1.3 ≤ Uout < 1.7

Line B: d = −12.5(Uout − 1.3)+20

to emerge in the workspace of the end-effector during oper- 

0.9 ≤ Uout < 1.3

Line C: d = −25.2(Uout − 0.9)+30

ation in case of accidents. The designed IOAM functional
module can achieve the RPM of obstacles and send the mon-  Line D: d = −50.5(Uout − 0.7)+40 0.7 ≤ Uout < 0.9
0.5 ≤ Uout < 0.7

Line E: d = −80.5(Uout − 0.5)+60

itoring data to the host computer. 

Uout < 0.5

Line F: d ≈ 80 → No obstacle

The IOA sensor (GP2Y0A21) is adopted to identify
whether there is obstacle in front of the sensor. The effective (1)

5544 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

TABLE 1. Main configuration parameters of the testbed.

a pilot-controlled check valve and a direct-acting proportional

FIGURE 7. Correspondence between output voltage and sensing distance. valve. During the experiments, the lifting speed of the cylin-
der is adjusted by controlling the opening size of the propor-
tional valve with a proportional amplifier. Two servomotors
are installed on each mobile module for the rotation of the
mobile module around the circular orbit and the rotation of
the cable drum.
A rectangular obstacle is located on the platform. The cable
drum is utilized for the wrapping of the cable through a fix
pulley to elevate the end-effector. Main configuration param-
eters of the testbed is given in Table 1. A specific illustration
of the connection mode between cables and the end-effector
is presented in Fig. 10 for both RPM and IOAM experiments.
FIGURE 8. Layout of infrared sensors and display panels. In Fig. 10(a), three connection joints B2, B4, B6 are dis-
tributed evenly on the top surface of the end-effector with a
distribution radius of Ru = 75mm.
According to the data packet from IOA sensors, the obsta- Meanwhile, the other three connection joints are dis-
cle information at a specific node is obtained after converting tributed evenly on the medial diagonal plane of the side circle
output voltage into sensing distance. The layout of infrared with a distribution radius of Rd = 125mm. The rotational
sensors and display panels are depicted in Fig. 8. It shows angle between cable 2 and cable 1 is π/6. As for the connec-
that eight sensors are arranged uniformly along the orbit. tion mode for IOAM experiment, three cables, numbered 1, 3,
In addition, four circular display panels are positioned along 5, are distributed evenly along circumferential direction and
the radial direction for each sensor, corresponding respec- they are all connected to a point B which locates just above
tively to four regions of 0∼15cm, 15∼30cm, 30∼45cm and the center of the top surface of the end-effector.
45∼60cm. Once the distance between the obstacle and the Referring to the hardware design of monitoring system in
node reduces to a certain value, the corresponding panel Section IV, the monitoring nodes are divided into IOAM node
changes its color immediately which denotes obstacles in the and RPM node. A picture that illustrates the installation of
area. these monitoring nodes is presented in Fig. 11.
The RPM functional module is to acquire the real-time 1) IOAM node. An infrared sensor GP2Y0A21 is installed
pose of the end-effector through the aforementioned forward at the IOAM node which is fixed on the outside surface of the
kinematic solution of the monitoring data from six displace- mobile module. Additionally, the sensor is installed right in
ment sensors. The theoretical trajectory of the end-effector front of the workspace.
displays in the host computer interface for the convenient 2) RPM node. A displacement sensor MPS-S-1000 is
check and comparison by operators. Any program that sat- installed at the RPM node, specifically in the front side of
isfies MATLAB language specifications can be written into the pulley. The sensor wire, fixed by a U-shaped plate and a
the internal of the script node. Then, the motion mode plan- pulley bracket, is guaranteed to be parallel to the cable. The
ning is accomplished through definitions of input and output signal output port of the sensor is accessed to the RPM node
parameters. through a wire pinboard.

IV. EXPERIMENTAL TESTBED V. PERFORMANCE EVALUATION AND DISCUSSION

Fig. 9 shows physically the experimental testbed for the A. REAL-TIME POSE MONITORING
MCDPR. A hydraulic pump with a supply power of only 1) FORWARD KINEMATIC SOLUTION
1MPa is used for the stretch of the hydraulic cylinder. The forward kinematic solution of MCDPR refers to the
The outlet of the pump is connected to six valve blocks solution of the spatial pose for end-effector with known
and each block is connected in series with a cut-off valve, cable length changes. Fig. 12 illustrates the schematic of a

VOLUME 7, 2019 5545

 S. Qian et al.: Development of MCDPR Systems

FIGURE 9. Picture of the experimental testbed for MCDPR.

FIGURE 10. Connection modes between cables and end-effector

for (a) RPM experiment; (b) IOAM experiment.

FIGURE 12. Schematic of a 6 cable-driven parallel mechanism.

parallel robot [31]. However, in order to ensure the real-time

FIGURE 11. Installation of monitoring nodes: (a) IOAM node, (b) RPM
node.
performance, the flexibility of the cable is neglected for
simplicity, which is treated as a massless body.
6 cable-driven parallel mechanism. The real-time end- The Jacobian matrix describes the linear relation between
effector pose is regulated by changing the cable lengths the velocities of six cables and end-effector [32], which can
through the control of corresponding motors, which can be expressed as
subsequently realize a high-accuracy spatial motion of the L̇ = J · Ṡ (2)
end-effector. Two coordinate systems, the global coordinate
system OXYZ and the local coordinate system O1 X1 Y1 Z1 , Ṡ is the end-effector velocity, which can be presented as
are established before the derivation of kinematic equations.
As depicted in Fig. 12, the origin of OXYZ locates in the Ṡ = [ ẋ ẏ ż α̇ β̇ γ̇ ]T (3)
orbit center on the ground while that of O1 X1 Y1 Z1 in the end-
L̇ is the velocities of six cables given by the representation
effector center. The end-effector has six DOFs with its spatial
T
pose described by the origin coordinates of O1 X1 Y1 Z1 (x, y,

L̇ = L̇1 L̇2 L̇3 L̇4 L̇5 L̇6 (4)
z), and the rotation angles of the end-effector around X , Y
and Z axes (α, β, γ ). In which, x, y, z separately denote X - J refers to a 6 × 6 Jacobian matrix that relates to the spatial
coordinate, Y -coordinate and Z -coordinate, and α, β, γ are velocity of the end-effector
deflection angle, pitch angle and roll angle, respectively.
∂Li ∂Li ∂Li ∂Li ∂Li ∂Li T
 
The sagging of cable should be taken into account J= (5)
when establishing the accurate model of the cable ∂x ∂y ∂z ∂α ∂β ∂γ

5546 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

Following variables are defined for simplification end-effector trajectory is defined as the following equations
 
∧ ∧ x = 0.25 − 0.25cos(πt 2)
(s = sin, c = cos) 

y = 0.25 sin(πt 2)
 
Xi1 = xAi − xBi



z = 0.1t+ 0.75

Yi1 = yAi − yBi (6) (0 ≤ t ≤ 4) (11)

Zi1 = zAi − zBi  α = πt 24
β = πt 24


 
Xi2 = −cβsγ x1Bi − cβcγ y1Bi

γ = πt 48

(7)

Xi3 = −sβcγ x1Bi + sβsγ y1Bi + cβz1Bi

 Yi2 = (cβsγ − sαsβSγ )x1Bi − (cβsγ + sαsβcγ )y1Bi where t represents time.

Yi3 = sαcβcγ x1Bi − sαcβsγ y1Bi + sαsβz1Bi

(8)

 Yi4 = (cαsβcγ − sαsγ )x1Bi
− (cαcγ + cαsβsγ )y1Bi − cαcβz1Bi



 Zi2 = (sαcγ + cαsβsγ )x1Bi − (sαsγ − cαsβcγ )y1Bi
Zi3 = cαcβcγ x1Bi + cαcβsγ y1Bi − cαsβz1Bi

(9)

 Zi4 = (cαsγ + sαsβcγ )x1Bi
+ (cαcγ − sαsβsγ )y1Bi − sαcβz1Bi


where (xAi , yAi , zAi ) and (xBi , yBi , zBi ) are the coordinates of
Ai and Bi in the global coordinate system OXYZ, (x1Bi , y1Bi ,
z1Bi ) are the coordinates of Bi in the local coordinate system
O1 X1 Y1 Z1 .
Then, the expression of each element in the kinematic
FIGURE 14. Desired trajectory and forward solution trajectory for
Jacobian matrix is derived as end-effector.
∂Li /∂x = Xi1 /Li


∂Li /∂y = Yi1 /Li A comparison of desired trajectory and forward solution




∂Li /∂z = Zi1 /Li trajectory is obtained as illustrated in Fig. 14. The end-


(10) effector trajectory is donated to be a rising spiral line. More-
 ∂Li /∂α = (Xi1 Xi2 + Yi1 Yi2 + Zi1 Zi2 ) /Li
∂Li /∂β = (Xi1 Xi3 + Yi1 Yi3 + Zi1 Zi3 ) /Li over, the changing trends of two trajectories are essentially




∂Li /∂γ = (Yi1 Yi4 + Zi1 Zi4 ) /Li coincident and the max range of error is 0.005m, which in


return validates the feasibility of utilizing the numeric-based
The end effector moves along the trajectory from initial solution algorithm for the forward kinematic solution of
pose in the workspace through changing the cable length. MCDPR.
The real-time length vector of each cable is symbolized
as L. Theoretically, P 0 , the initial pose vector of the end- 2) RPM EXPERIMENTS
effector, is possible to choose any spatial pose. In this study, In RPM experiments, the motion mode is assigned to be ver-
the forward kinematic solution of MCDPR can be simplified tical lifting motion. X, Y and Z displacements are separately
to the current pose solution of the end-effector according 0m, 0.25m and 0m. Deflection angles of three axes are all 0.
to P 0 and L. Fig. 13 summarizes the flow chart of the Fig. 15 displays a comparison of theoretical length and actual
numeric-based algorithm for forward pose solution. length for six cables. The variations of cable length with time
are presented to be substantial agreement with each other for
six cables. Meantime, a slight fluctuation is still found in
the measured curves due to the inadequate system stiffness,
along with the vibration induced by flexible characteristics
associated with cables.
Referring to the comparison of theoretical and measured
trajectory shown in Fig. 16, when the end effector moving in
the Z direction, the measured trajectory fits the theoretical
trajectory with good accuracy (see Fig.16 (c)). However,
the increased tension and the resulted elastic extension of
cables during the movement cause wild movements of the val-
ues in X direction, Y direction and three pose angles, includ-
ing deflection angle, pitch angle and roll angle. A detailed
FIGURE 13. Numeric-based algorithm for forward pose solution. illustration of the pose deviation for the end-effector is given
in Table 2.
On the basis of the algorithm shown in Fig. 13, kine- As for the absolute displacement deviations in X, Y and
matics is simulated using MATLAB software. The desired Z directions, the maximum values are 0.03959m, 0.04544m

VOLUME 7, 2019 5547

 S. Qian et al.: Development of MCDPR Systems

FIGURE 15. Theoretical length and actual length for six cables: (a) cable 1, (b) cable 2, (c) cable 3, (d) cable 4, (e) cable 5, (f) cable 6.

TABLE 2. Pose deviation of the end-effector. shown in Fig. 2a, which is widely used in engineering appli-
cations, is chosen for the cooperative operation analysis.
The schematic of the MCDPR for configuration I is shown
in Fig. 17.
The end-effector is simplified to be a mass point with its
coordinates to be (XB , YB , ZB ) in the global coordinate system
OXYZ. Then, the coordinate vector of the end point B are
described by
B = (XB , YB , ZB ) (12)
The coordinate vector for three top ends A1-A3 is given as
and 0.00796m while the average values are only 0.01624m, Ai = (Ro cos (θi ) , Ro sin (θi ) , hi ) (i = 1, 2, 3) (13)
0.01719m, respectively. Furthermore, the average deviations
of deflection angle, pitch angle and roll angle are 1.091o, where θi is the rotational angle of the ith mobile module.
1.628o and 3.954o, respectively. The end-effector trajectory Hence, based on the inverse kinematic model in [3]. The
tracking error is caused by different error sources, including length of ith cable is derived as
q
kinematic error, deformation and servo error. Kinematic error
Li = (XB − Ro cos (θi ))2 +(YB − Ro sin (θi ))2 +(ZB − hi )2
caused by machining, assembly and operation is the major
error sources, including the influence of force sensors and (i = 1, 2, 3 ) (14)
pulleys on the kinematics [33]–[35]. Note that these devia- Due to the configuration, the static mechanical model of
tions can be eliminated by the purposed robust iterative learn- the end-effector is expressed by
ing controller in [9]. Thus, the end-effector can be elevated 3 −→
smoothly within the accepted error range. X Ai B
Fi · −→ = Me g (15)
Ai B

i=1
B. OBSTACLE AVOIDANCE
1) THEORETICAL MODEL
where Fi is the tension force of ith cable. for other configura-
tions, static mechanical model can be derived similarly.
As seen in Fig. 2, the RCDPR can change their module and
Since the cable can only provide tension force and it is
size to meet specific operational demands for the simula-
guaranteed not to be loosened during operation, then
tion of the task-oriented cooperative work of 3 to 6 mobile
modules. In this study, the 3 modules type configuration Fmax ≥ Fi ≥ Fmin (16)

5548 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

FIGURE 16. Theoretical trajectory and measured trajectory for the end-effector: (a) X-displacement, (b) Y-displacement, (c) Z-displacement,
(d) deflection angle, (e) pitch angle, (f) roll angle.

FIGURE 17. Schematic of the MCDPR for configuration I.

where Fmax and Fmin are respectively the maximum and

minimum tension forces.
Based on the aforementioned static mechanical model of
the MCDPR (configuration I), the workspace calculation
is carried out with following parameters given as Fmax =
Me g, Fmin = 0. The results for various cylinder heights hi
FIGURE 18. Workspaces for various cylinder heights and rotational
and rotational angles θi are obtained as depicted in Fig. 18. angles: (a).
The workspace can be approximately regarded as a triangle
formed by three mobile modules. In addition, the volumes of
the workspaces are calculated as 0.1293, 0.1713, 0.0882 and The environment is treated as a potential field where the
0.1138m3 for different conditions shown in Fig. 18(a) to (d), combinations of attraction and repulsion elements are used
respectively. to represent the destination and obstacles, respectively. Con-
Fig. 19 is the simplified top view of the RCDPR, where tinuous and smooth motion is obtained with closed-form
the gray polygons represent the workspace formed by sev- expressions of the trajectory velocity. During the end-effector
eral mobile modules. Both the end-effector and cables moving along the given trajectory, all these cables need to
need to avoid collisions with obstacles during operation. avoid collisions with obstacles. As three mobile modules
For the end-effector, its trajectory can easily obtained by locates at the initial position A1-A2-A3 and the end-effector
using the artificial potential field method, which is widely moves from the start point S to the goal point G, cable 3
adopted for obstacle avoidance in engineering practice. (line PA3) is inevitably collide with the obstacle. To prevent

VOLUME 7, 2019 5549

 S. Qian et al.: Development of MCDPR Systems

FIGURE 19. Sketch of the obstacle avoidance process. FIGURE 21. Abrupt change positions of boundary points.

this, three mobile modules can be reconfigured to the position

A10 -A20 -A30 . Now, both the cables and end-effector are able
to avoid collisions with the obstacle
To identify the position to ensure each mobile modules
can move without collisions, the anchor point of each cable
is assumed to be a ‘‘light source’’. And the arc that formed
by the cast shadow of the obstacle is perceived as the for-
bidden area of the corresponding mobile module. As shown
in Fig.20, the anchor point of each cable is coincide with each
other, so the read dashed arc is the common forbidden area of
these three mobile module. For other type of configurations,
the forbidden area of each mobile module need to be calcu-
lated, respectively.
FIGURE 22. Real-time display interface of IOAM.

FIGURE 20. Schematic of the obstacle avoidance algorithm.

In Fig. 20, r represents the distance between an arbitrary

point P on the trajectory and the origin point O. E and F are
two boundary points of the obstacle projection on the orbit FIGURE 23. Simulation results at four abrupt change positions: (a) I1 ;
(b) I2 ; (c) I3 ; (d) I4 .
closed to mobile module 1 and mobile module 2, respectively.
Q is the intersection point between the extended line PO and
the formula below
the orbit. Here, 6 POX = ϕ, 6 EPQ = ψ1 , 6 PEO = ψ2 ,
6 EOQ = α1 , 6 QOF = α2 . Once the coordinate of P is given, cos ψ1 = b2+ r 2 − R2o 2br



the value of ψ1 and ϕ can be calculated using the coordinates ψ2 = arccos b2 + R2o − r 2 2bRo

 
(17)
of P and obstacle corners. Then, α1 can be calculated using α1 = ψ1 + ψ2


5550 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

FIGURE 24. A comparison of simulation results and experimental data: (a) rotational angle; (b) cable length; (c) force.

where b is the distance between the arbitrary point P and Simulation and experiment of obstacle-avoidance for the
boundary point E. The value of α2 can be obtained using the MCDPR are conducted sequentially with the parameters
above approach similarly. below: Initial end-effector position is (−0.3m, 0, 0.25m), ter-
The rotational angles between line OE, line OF and X -axis minal end-effector position is (0.3m, 0, 0.25m), and cylinder
are symbolized as τ1 , and τ2 with their values obtained as height are h1 = h2 = h3 = 0.66m. The values of rotational
τ1 = ϕ + π − α1 and τ2 = ϕ + π + α2 . Then, the positions angles θ1 , θ2 , θ3 are solved by substituting the end-effector
of two boundary points of the inferior arc can also be solved. trajectory into (17) and (18), which obtained via the artificial
Here a buffer angle η is employed to reduce the risk of potential field method [36]. The movement process of the
collision between cables and the obstacle, the position angles MCDPR utilized the obstacle avoidance algorithm is sim-
θi between mobile module 1, 2, 3 and X -axis are set as ulated and the results at four abrupt change positions are
displayed in Fig. 23. The results indicate that the MCDPR
 θ 1 = τ1 − η

can effectively achieve obstacle-avoidance after reconfigura-
θ 2 = τ2 + η (18) tion by rapidly changing the circumferential position of the
θ3 = θ1 + θ2

mobile module at abrupt change positions.
Note that the position angles θ3 is set to be the average Fig. 24 shows the variations of rotational angle, cable
value of θ1 and θ2 to make cable force evenly distributed. For length and force with time of simulation and experiment.
other type of configurations, the position angles θi of each In the figure, the solid line donates the experimental data
mobile module can be obtained similarly. while the dot line represents the simulation results. Two
Theoretically, an abrupt change in the positions of bound- inflection points are found on the curves of both θ1 and θ2 ,
ary points E, F may appear when the end-effector pass corresponding separately to the moment of four abrupt
over the intersection points between the trajectory and the change positions shown in Fig. 23. In specific, mobile mod-
extensions of obstacle edges. As shown in Fig. 21, when ules 1, 2 abruptly alter their circumferential positions at 170s,
the end-effector moves from P1 to I1 then to P2 , the pivot 355s, and 125s, 310s. Moreover, the curves of θ1 and θ2 are
of boundary line PF switches from C2 to C1 , causing an symmetric due to the symmetry of the rectangular obstacle
abrupt change in the motion gain of the end-effector. More- and the interchangeability between mobile modules 1 and 2.
over, the boundary line between line PQ and the movement Similarly, the symmetry and inflection points also appear in
direction of the end-effector switches its pivot firstly at each the curves of θ3 and subsequent cable length Li and force Fi .
obstacle corner. In the figure, I1 and I3 are the abrupt change In addition, the variation trend of experimental results are in
positions of F, I2 and I4 are the abrupt change positions of good agreement with the simulation values. However, due
E. Note that a constant safe angle is used to avoid collision, to the kinematic error sources, measurement error and the
as seen in (18), the abrupt change of boundary points will acceleration neglected in the static mechanical model, there
directly reflect in the movement of the mobile module. are some differences between the results of experiment and
simulation, particularly at inflection points.
2) SIMULATION AND EXPERIMENT OF IOAM
The IOAM node can monitor the real-time position of obsta- VI. CONCLUSIONS
cles placed arbitrarily in the workspace. The display interface In this paper, design, safety monitoring, and obstacle
of IOAM is plotted in Fig. 22. In the figure, the obstacle distri- avoidance for the MCDPR is presented to meet the
bution sub-interface is intended for the real-time visualization performance requirements in terms of multifunction and
of the approximate position of obstacles. Simultaneously, payload-to-weight ratio for the CDPRs. Modular design and
the exact obstacle position in the radial direction is available system description of the MCDPR were presented. The paper
from the obstacle-node distance sub-interface. Note that the detailed the hardware and software implementations of the
sampling period is 0.01s in the experiments. system. The kinematics of the MCDPR is analyzed using

VOLUME 7, 2019 5551

 S. Qian et al.: Development of MCDPR Systems

forward solution method and the theoretical model for obsta- [13] Y. Mao and S. K. Agrawal, ‘‘Design of a cable-driven arm exoskele-
cle avoidance of the MCDPR with three DOFs was derived. ton (CAREX) for NEURAL rehabilitation,’’ IEEE Trans. Robot., vol. 28,
no. 4, pp. 922–931, Aug. 2012.
Two categories of experiments were conducted, including [14] P. K. Jamwal, S. Q. Xie, S. Hussain, and J. G. Parsons, ‘‘An adaptive
the real-time pose monitoring and obstacle avoidance. For wearable parallel robot for the treatment of ankle injuries,’’ IEEE/ASME
the pose deviation of the end-effector, the mean values were Trans. Mechatronics, vol. 19, no. 1, pp. 64–75, Feb. 2014.
[15] B. Zi, H. Sun, and D. Zhang, ‘‘Design, analysis and control of a winding
0.01624m and 0.01719m in X and Y directions while that hybrid-driven cable parallel manipulator,’’ Robot. Comput.-Integr. Manuf.,
in Z direction was rather small, only 0.00796m. Simultane- vol. 48, pp. 196–208, Dec. 2017.
ously, the average deviations of deflection angle, pinch angle [16] B. Zi, J. Lin, and S. Qian, ‘‘Localization, obstacle avoidance planning and
control of a cooperative cable parallel robot for multiple mobile cranes,’’
and roll angle were 1.091◦ , 1.628◦ and 3.954◦ , respectively. Robot. Comput.-Integr. Manuf., vol. 34, pp. 105–123, Aug. 2015.
In the obstacle avoidance experiments, mobile modules 1, [17] S. Qian, B. Zi, W.-W. Shang, and Q. Xu, ‘‘A review on cable-driven parallel
2 abruptly alter their circumferential positions at 170s, 355s, robots,’’ Chin. J. Mech. Eng., vol. 31, p. 66, Dec. 2018.
[18] G. Rosati, D. Zanotto, and S. K. Agrawal, ‘‘On the design of adaptive
and 125s, 310s, which leads to corresponding inflections cable-driven systems,’’ J. Mech. Robot., vol. 3, no. 2, p. 021004, 2011.
of rotational angle, cable length and force. Additionally, [19] X. Zhou, C. P. Tang, and V. Krovi, ‘‘Analysis framework for cooperating
the curves of cable length and force appear symmetrically mobile cable robots,’’ in Proc. IEEE Int. Conf. Robot. Automat., vol. 20,
May 2012, pp. 3128–3133.
with respect to the time midpoint due to the symmetry of [20] L. Gagliardini, S. Caro, M. Gouttefarde, and A. Girin, ‘‘Discrete recon-
the rectangular obstacle and the interchangeability between figuration planning for cable-driven parallel robots,’’ Mechanism Mach.
mobile modules. Theory, vol. 100, pp. 313–337, Jun. 2016.
[21] S. Abdolshah, D. Zanotto, G. Rosati, and S. K. Agrawal, ‘‘Optimizing
Experimental results confirmed the feasibility and effi- stiffness and dexterity of planar adaptive cable-driven parallel robots,’’
ciency of the kinematics model and obstacle avoidance J. Mech. Robot., vol. 9, no. 3, p. 031004, 2017.
method. The motion control and safety monitoring system [22] J. Lin, C. S. Huang, and J. Chang, ‘‘A mechatronic kit with a control
methodology for a modualized cable-suspended robot,’’ J. Vib. Control,
were proved to be effective for realization of real-time safety vol. 22, no. 20, pp. 4211–4226, 2016.
monitoring and obstacle avoidance of the MCDPR. Future [23] G. Castelli and E. Ottaviano, ‘‘Modelling and simulation of a cable-based
work will focus on efficient reconfiguration strategies of parallel manipulator as an assisting device,’’ in Computational Kinematics.
Berlin, Germany: Springer, 2006, pp. 17–24.
MCDPR for better performance and flexibility in cluttered [24] M. Anson, A. Alamdari, and V. Krovi, ‘‘Orientation workspace and stiff-
environment, by determining the optimal number of the ness optimization of cable-driven parallel manipulators with base mobil-
mobile modules and locations of the cable anchor points. ity,’’ J. Mech. Robot., vol. 9, no. 3, p. 031011, 2017.
[25] G. Abbasnejad, J. Yoon, and H. Lee, ‘‘Optimum kinematic design of a
REFERENCES planar cable-driven parallel robot with wrench-closure gait trajectory,’’
Mechanism Mach. Theory, vol. 99, pp. 1–18, May 2016.
[1] S. H. Yeo, G. Yang, and W. B. Lim, ‘‘Design and analysis of cable-driven
[26] A. Peidró, A. Gil, J. M. Marín, and Ó. Reinoso, ‘‘A Web-based tool to
manipulators with variable stiffness,’’ Mechanism Mach. Theory, vol. 69,
analyze the kinematics and singularities of parallel robots,’’ J. Intell. Robot.
pp. 230–244, Nov. 2013.
Syst., vol. 81, no. 1, pp. 145–163, 2016.
[2] B. Zi and S. Qian, Design, Analysis and Control of Cable-Suspended
[27] B. S. Dhillon, A. R. M. Fashandi, and K. L. Liu, ‘‘Robot systems reli-
Parallel Robots and Its Applications. Singapore: Springer, 2017.
ability and safety: A review,’’ J. Qual. Maintenance Eng., vol. 8, no. 3,
[3] B. Zi and Y. Li, ‘‘Conclusions in theory and practice for advancing the pp. 170–212, 2002.
applications of cable-driven mechanisms,’’ Chin. J. Mech. Eng., vol. 30, [28] J. T. C. Tan, F. Duan, R. Kato, and T. Arai, ‘‘Safety strategy for
no. 4, pp. 763–765, 2017. human–robot collaboration: Design and development in cellular manufac-
[4] Y. Wang, G. Yang, T. Zheng, K. Yang, and D. Lau, ‘‘Force-closure turing,’’ Adv. Robot., vol. 24, nos. 5–6, pp. 839–860, 2010.
workspace analysis for modular cable-driven manipulators with co-shared [29] E. Cabal-Yepez, A. G. Garcia-Ramirez, R. J. Romero-Troncoso,
driving cables,’’ in Proc. 13th IEEE Conf. Ind. Electron. Appl. (ICIEA), A. Garcia-Perez, and R. A. Osornio-Rios, ‘‘Reconfigurable monitoring
May/Jun. 2018, pp. 1504–1509. system for time-frequency analysis on industrial equipment through
[5] C. Gosselin and S. Foucault, ‘‘Dynamic point-to-point trajectory plan- STFT and DWT,’’ IEEE Trans. Ind. Informat., vol. 9, no. 2, pp. 760–771,
ning of a two-DOF cable-suspended parallel robot,’’ IEEE Trans. Robot., May 2013.
vol. 30, no. 3, pp. 728–736, Jun. 2014. [30] N. Sun, Y. Fang, and H. Chen, ‘‘A continuous robust antiswing track-
[6] B. Y. Duan, Y. Y. Qiu, F. S. Zhang, and B. Zi, ‘‘On design and experiment ing control scheme for underactuated crane systems with experimental
of the feed cable-suspended structure for super antenna,’’ Mechatronics, verification,’’ J. Dyn. Syst., Meas., Control, vol. 138, no. 4, p. 041002,
vol. 19, no. 4, pp. 503–509, 2009. 2016.
[7] B. Zi, N. Wang, S. Qian, and K. Bao, ‘‘Design, stiffness analysis and [31] J.-P. Merlet and J. Alexandre-Dit-Sandretto, ‘‘The forward kinematics of
experimental study of a cable-driven parallel 3D printer,’’ Mechanism cable-driven parallel robots with sagging cables,’’ in Cable-Driven Parallel
Mach. Theory, vol. 132, pp. 207–222, Feb. 2019. Robots. Cham, Switzerland: Springer, 2015, pp. 3–15.
[8] S. P. Donohoe, S. A. Velinsky, and T. A. Lasky, ‘‘Mechatronic implementa- [32] M. Carricato and J.-P. Merlet, ‘‘Stability analysis of underconstrained
tion of a force optimal underconstrained planar cable robot,’’ IEEE/ASME cable-driven parallel robots,’’ IEEE Trans. Robot., vol. 29, no. 1,
Trans. Mechatronics, vol. 21, no. 1, pp. 69–78, Feb. 2016. pp. 288–296, Feb. 2013.
[9] S. Qian, B. Zi, and H. Ding, ‘‘Dynamics and trajectory tracking control of [33] E. Ottaviano, M. Ceccarelli, and F. Palmucci, ‘‘An application of CaTraSys,
cooperative multiple mobile cranes,’’ Nonlinear Dyn., vol. 83, nos. 1–2, a cable-based parallel measuring system for an experimental characteriza-
pp. 89–108, Jan. 2016. tion of human walking,’’ Robotica, vol. 28, no. 1, pp. 119–133, 2010.
[10] K. H. Cho et al., ‘‘Inspection robot for hanger cable of suspension [34] A. Pott, ‘‘Influence of pulley kinematics on cable-driven parallel robots,’’
bridge: Mechanism design and analysis,’’ IEEE/ASME Trans. Mechatron- Latest Advances in Robot Kinematics. Dordrecht, The Netherlands:
ics, vol. 18, no. 6, pp. 1665–1674, Dec. 2013. Springer, 2012, pp. 197–204.
[11] Q. Chen, W. Chen, G. Yang, and R. Liu, ‘‘An integrated two-level self- [35] A. Gonzalez-Rodriguez, F. J. Castillo-Garcia, E. Ottaviano, P. Rea, and
calibration method for a cable-driven humanoid arm,’’ IEEE Trans. Autom. A. G. Gonzalez-Rodriguez, ‘‘On the effects of the design of cable-Driven
Sci. Eng., vol. 10, no. 2, pp. 380–391, Apr. 2013. robots on kinematics and dynamics models accuracy,’’ Mechatronics,
[12] H. Kino, S. Kikuchi, Y. Matsutani, K. Tahara, and T. Nishiyama, ‘‘Numeri- vol. 43, pp. 18–27, May 2017.
cal analysis of feedforward position control for non-pulley musculoskeletal [36] D. Lau, J. Eden, and D. Oetomo, ‘‘Fluid motion planner for nonholo-
system: A case study of muscular arrangements of a two-link planar system nomic 3-D mobile robots with kinematic constraints,’’ IEEE Trans. Robot.,
with six muscles,’’ Adv. Robot., vol. 27, no. 16, pp. 1235–1248, 2013. vol. 31, no. 6, pp. 1537–1547, Dec. 2015.

5552 VOLUME 7, 2019

S. Qian et al.: Development of MCDPR Systems

SEN QIAN received the Ph.D. degree from the DAOMING WANG received the Ph.D. degree
China University of Mining and Technology, from the China University of Mining and Technol-
China, in 2015. He is currently a Lecturer with the ogy, China, in 2014. He is currently an Associate
School of Mechanical Engineering, Hefei Univer- Professor with the School of Mechanical Engi-
sity of Technology, China. His research interests neering, Hefei University of Technology, China.
include robotics and automation. His research interests include mechatronics and
smart materials.

BIN ZI received the Ph.D. degree from Xidian YUAN LI is currently pursuing the Ph.D. degree
University, China, in 2007. He is currently a with the School of Mechanical Engineering, Hefei
Professor, the Dean of the School of Mechani- University of Technology, China.
cal Engineering, and the Director of the Robotics
Institute, Hefei University of Technology, China.
His research interests include robotics and automa-
tion, mechatronics, and multirobot systems.

VOLUME 7, 2019 5553