See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/366570514

Enhanced Routing Technique for Military Dog Robot

Preprint · December 2022

CITATIONS READS

0 14

16 authors, including:

Fatma M. Talaat
Kafrelsheikh University
17 PUBLICATIONS 201 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Effective Load Balancing Strategy for Real‑Time Fog Computing Environment View project

All content following this page was uploaded by Fatma M. Talaat on 24 December 2022.

The user has requested enhancement of the downloaded file.

 Enhanced Routing Technique for Military Dog Robot
Fatma M. Talaat1, Ahmed Reda Mohamed Abd-Elnaem2, Hagar Emad El-dein Abd-Elrahman2, Diaa m.
Zayan2, Mayar Mohamed AbdElaty2, Ahmed khaled Fawzy Fawzy2, Sohib Ebrahim Mohamed Mansour2,
Raneem Emad Mohamed2, Mostafa Kamel Ali Muhammad2, AbdEL-Rahman Kasem Muhammad Ali2,
Mahmoud Ashraf Amer2
[email protected]
1Machine Learning & Information Retrieval Department, Faculty of Artificial Intelligence, Kafrelsheikh University, Kafrelsheikh, Egypt
2Faculty of Artificial intelligence,kafrelsheikh,Egypt

ORCID: 0000-0001-6116-2191

Abstract
This research suggests a new robotic system technique that works specifically in settings like hospitals or
emergency situations when prompt action and the preservation of human life are crucial. The goal of the proposed
work is to leverage technology to force-multiply the CBP presence, as well as reduce human exposure to life-
threatening hazards. The suggested Route Planning Algorithm (RPA) based on Reinforcement Learning offers
effective services by gathering and sending data between robots and human while humans are kept out of the
danger. There are three key modules that make up RPA: (i) Robot Finding Module (RFM), (ii) Robot Charging
Module (RCM), and (iii) Route Selection Module (RSM). Using such autonomous systems like RPA in places where
there is a need for human gathering, is very essential particularly in the medical field reducing the risk of virus
spread, which can save thousands of lives. The simulation results using the proposed framework show the flexible
and efficient movement of the robots compared to conventional methods under various environments. RSM is
contrasted with the leading cutting-edge topology routing options. RSM's primary benefit is the much reduced
calculations and updating of routing tables. In contrast to earlier algorithms, RSM produces the lowest AQD.
RSM is hence an appropriate algorithm for real-time systems.
Keywords: Autonomous Robots, Routing Algorithm, Collision Avoidance, Reinforcement Learning.
1. Introduction
For communication and collaboration among linked smart devices, current smart applications and
settings (such as smart homes, intelligent industry systems, and healthcare systems) significantly rely on
the cloud. While cloud servers and datacenters are primarily centralized, smart devices are dispersed. In
light of this, a critical issue is the response time of data transmission between the cloud and smart devices,
especially for applications that have strict delay requirements, such as saving lives in emergencies.
Although the entire Fog platform has not yet been fully developed, new technology paradigms for Fog
computing have recently been proposed to tackle this issue.
By bringing processing power from the distant Cloud closer to the end users, the Fog platform's major
objective is to increase end users' access to real-time interaction and location-based services. The Fog's
on-site processing capacity greatly lessens the Cloud's data load. There are many mobile robots available
today that can operate alongside people or possibly take the place of humans in the majority of
occupations. They may become involved in people's lives because they offer comfort. Because they may
lower the quantity of labor needed in manufacturing, they use them in both their daily lives and the bulk
of industrial fields. They can also perform tasks that are impossible for people to perform or save lives
in hospitals or other emergency situations.
A new generation of coronavirus pneumonia had broken out by the middle of December 2019 in Wuhan,
China, drawing the attention of people all over the world [1] [2] [3]. On January 30, 2020, the World
Health Organization (WHO) deemed the worldwide COVID-19 outbreak a public health emergency of
international concern [4]. As COVID-19 infections increase, more healthcare workers are getting sick—
up to 10% of them in some countries [7]. The precise number of doctors, nurses, and other healthcare

1
workers who have tested positive is still unknown, despite the fact that the number of COVID-19 cases
worldwide is still being regularly tracked [8] [9]. This is important information because these individuals
are most at risk of contracting the virus and dying as a result.
It may seem irrational to be concerned about the increasing death toll among healthcare professionals,
but frontline healthcare workers are dying in China [10] and Italy [11]. At least 200 nurses in the USA
have been suspended from work and placed in isolation due to a lack of testing, while in Pittsfield,
Massachusetts, more than 160 members of the staff of Berkshire Medical Center have been quarantined
due to exposure to the virus, and in Washington, scores of medical staff at a nursing center have tested
positive for COVID-19. "At least 2,629 healthcare workers have COVID-19, accounting for about 8.3
percent of all cases in Italy."
The goal of the proposed work is to leverage technology to force-multiply the CBP presence, as well as
reduce human exposure to life-threatening hazards. The suggested Route Planning Algorithm (RPA)
based on Reinforcement Learning offers effective services by gathering and sending data between robots
and human while humans are kept out of the danger.
There are three key modules that make up RPA: (i) Robot Finding Module (RFM), (ii) Robot Charging
Module (RCM), and (iii) Route Selection Module (RSM). Using such autonomous systems like RPA in
places where there is a need for human gathering, is very essential particularly in the medical field
reducing the risk of virus spread, which can save thousands of lives.
The novelty in this work is based on building a complete autonomous robot system considering all criteria
in such systems like:
i. Selecting the optimal robot for the incoming request using new proposed equation which has the
needed requirements according to the problem issue.
ii. Finding the best path for the selected robot to move through using Tree-Based Path Finding
Algorithm.
iii. Building a new recharging algorithm to recharge the robots periodically when it’s mandatory to
save time and cost.
The proposed algorithms grunted that the system is reliable and available. Using RL here is not like
traditional usage. RL here is used in the controlling point not in the robot. RL has been used for several
needs before like in [16], authors used RL to propose a new effective hyperparammetrs optimization
algorithm for CNN.
The remaining work is organized as follows. In section 2, some of the recent related work in the routing
algorithms are presented. Section 3 presents the proposed framework. Experimental evaluation is
provided in section 4. And in section 5, we conclude this work.
2. Literature Review
Robotic dogs are robots that are designed to resemble dogs in appearance and behavior, usually
incorporating dog-like actions such as barking, tail-wagging, and following commands. This section
introduces a collection of Robotic Dog projects from the past as well as some techniques that were used
in their creation.
J. Zico Kotler, Andrew Y. Ng [12] presented a software system which allows a quadruped robot to
maneuver through a wide variety of challenging terrain quickly and reliably, using static and dynamic
modes of locomotion. The software architecture made use of both static and dynamic gaits, as well as
specialized dynamic maneuvers. The central parts of their approach are; (i) the use of learning algorithms

2
to learn route planning, footstep planning, and dynamic maneuvers, and (ii) A focus on rapid recovery
and replanning to overcome situations where the robot deviates from its desired trajectory.
Marco Hutter, et al. [13] presented ANYmal, a quadrupedal robot that features outstanding mobility and
dynamic motion capability, thanks to novel, compliant joint modules. The robot is very robust against
impulsive loads during running or jumping, and it was designed with a focus on outdoor suitability,
simple maintenance, and user-friendly handling. ANYmal was considered a step towards combining high
mobility with dynamic locomotion capability. There was a special emphasis put on having a simple to
maintain system, which was achieved using the modular joint units “ANYdrive” that allow the creation
of robots of different kinematic structures very simply. The experiments carried out on the robot support
the claim of its robustness. Besides the improved protection, the biggest advantage of ANYmal is
certainly the outstanding range of motion in all joins, enabling a large variety of maneuvers to overcome
obstacles or to get up after falling, as well as simplified motion planning due to there being less internal
system constraints.
Gerardo Bledt, et al. [14] presented MIT Cheetah 3, a robust and dynamic quadruped robot. It utilizes
tailored mechanical design to enable simple control strategies for dynamic locomotion and features high-
bandwidth proprioceptive actuators to manage physical interaction with the environment. It presents a
new leg design that allows for more range of motion on the hips and knees. General balance and
locomotion controllers for Cheetah 3 are presented which allow the robot to handle unexpected terrain
disturbances through reactive gait modification. It demonstrates robustness even without the use of
external sensing, which signifies its capability of successful locomotion in challenging situations without
having prior knowledge of the environment to rely upon. The general control architecture uses any
number of different controllers with minimal modifications to the software or hardware. Promising initial
results of a novel nonlinear Policy Regularized Model Predictive Control Framework (PR-MPC).
Greedy best-first search algorithm always selects the path which appears best at that moment. It is the
combination of depth-first search and breadth-first search algorithms. It uses the heuristic function and
search. Best-first search allows us to take the advantages of both algorithms. With the help of best-first
search, at each step, we can choose the most promising node.
A* search is the most commonly known form of best-first search. It uses heuristic function h(n), and cost
to reach the node n from the start state g(n). In A* search algorithm, we use search heuristic as well as
the cost to reach the node. Hence we can combine both costs as following, and this sum is called as a
fitness number. An improved A* algorithm combined with the greedy algorithm is applied to multi-
objective point planning [15].
3. The proposed Route Planning Algorithm (RPA)
This section proposes a new Route Planning Algorithm (RPA). The robots are autonomous robots
sending signals with their location, dimensions, and charging limit to the controller server via sensors.
The controller controls the all movement of the robot. It selects the best robot to achieve a specific task,
detects the best path for the robot to move through, and decide the time of the charging for each robot.
There are three key modules that make up RPA as shown in Figure 1: (i) Robot Finding Module (RFM),
(ii) Robot Charging Module (RCM), and (iii) Route Selection Module (RSM).

3
 Cloud Layer

Backend Servers

Web Server Items Database

User Request

IOT Layer Fog Layer

Controller Robots
Database
Robots

Robot Robot Selection Robot

Charging
Routing Knowledge
Module

Ranking Rejection Items DB

Figure 1: Route Planning Algorithm (RPA)

3.1. Robot Finding Module (RFM)

In this module, the best robot for a specific task is selected according to its features and the requirements
of the task. RFM includes three main procedures: (i) Item Requests Receiving Algorithm (IRRA), and
(ii) Robot Rejection Space Algorithm (RRSA).
i. Item Request Receiving Algorithm (IRRA)
The Item Requests Receiving Algorithm (IRRA) is responsible for receiving the requests coming from
the user and collecting all the needed information about each request in a table called Request Information
Table (RIT). Table 1 illustrates an example of RIT table at a specific event. All database is stored in the
controller.
Table 1: Request Information Table (RIT)
Request Item Source Destination
User Item Code Area
Priority Weight Location Location

4
 User 713 Normal Item C-11 Fog C 0.1 KG 11,27 10,0

User 62 Critical Item A-1246 Fog A 6.3 KG 14,66 0,0

User 213 Low Item B-67 Fog B 1.7 KG 92,16 99,99

User 14 Normal Item A-112 Fog A 9.1 KG 97,14 0,0

User 19 Low Item D-4625 Fog D 6.5 KG 33,27 0,40

User 32 Normal Item A-17264 Fog A 0.9 KG 0.9 KG 0,0

ii. Robot Rejection Space Algorithm (RRSA)

The Robot Rejection Space Algorithm (RRSA), as shown in Figure 2, is responsible for determining
which robots will be excluded while selecting the best robot to the incoming request. The robots that will
be excluded are the robots with these specifications: (i) Busy robots; robots performing a task in the
current time. (ii) Charging robots: robots in their charging process in the current time. (iii) If the delivery
process has more time than the remaining charge time. (iv) If the item weight is much heavier than the
threshold weight that robot can hold.
Start a new Robot
Request to deliver item

Check Robot Status

from RFT

Yes No
Charging

Yes No
Busy

No Yes No
Priority=”Critical” TT to Deliver item
> Robot CT

Yes
Item Weight >
Robot TH
Yes

No

Join Robot
Ranking Algorithm
Reject Robot

Next Robot

5
 Figure 2: Robots Rejection Space Flowchart

3.2. Robot Charging Module (RCM)

The RCM is used to monitor and recharge robots periodically to keep the system available as much as
possible. The steps of RCM are shown in Algorithm 1.
Robot Charging Module (RCM
1. Sort all robots from the lowest charge to the highest
2. For each robot
3. If robot is not busy and robot charge below 60% and there is a free charger
4. Send robot to charge
5. Enf If
6. Next
Algorithm 1. Robot Charging Module (RCM)

3.3. Route Selection Module (RSM)

In order to get where robot needs to go, a robot has to have a physical path through the space, a need that
is often complicated by the presence of obstacles or other movement constraints. Tree-Based Path
Finding Algorithm uses a rapidly-exploring random tree, or RRT*, to find a path through a 2-dimensional
environment containing a number of differently-sized obstacles. The steps of the TPFA are shown in
Figure 3.

Figure 3. TPFA Steps

6
To grow the tree, when picking a random spot in the space, then figure out which of the existing nodes
is closest to the random point. Then, we draw a straight line between the random point and the closest
node, and create a new node along that line a certain distance away from the closest node. By choosing
a series of random points and creating a series of new nodes, the tree grows in different directions
throughout the space. A 200-node RRT* can be seen in Figure 4.

Figure 4. A 200-node RRT*

RRT* can be used to navigate around obstacles between two points. When generating each new node,
we perform collision detection with each obstacle, and we also note which existing node it’s connected
to - the “parent” node that I mentioned before. Another way of thinking about the parent is that it’s the
node in the tree that the randomly generated point was closest to. Keeping track of these parent nodes is
important because it allows us to re-trace our steps once we find a clear path to the end point. By jumping
from parent node to parent node all the way back to the starting point, we reconstruct the path that we
can tell the robot to follow
4. Implementation and Evaluation
The employed dataset, performance indicators, and evaluation findings were all reported in this part.
4.1. Dataset
We assume a grid map representing a group of robots as shown in Figure 5.

R1 R2

R4

R3
7
 Figure 5. An illustration of a grid layout that shows how the group of robots would avoid obstacles
4.2. Dog robot design
The steps for the designing of the dog robot are shown in Figure 6.

8
9
 Figure 6. Robot Design Steps

4.3. RSM Evaluation

The TPFA first uses the RRT* algorithm to find the best path which is the most suitable path in the
smallest time while avoiding the collision as shown in Figure 7. Then, it saves this path and uses it for
the robot movement.
This algorithm outperforms the other path finding algorithms as it is fast and it is also avoid any obstacle
without any collision.

Figure 7. Best Path Finding and Avoiding Collision

The performance metrics values are presented in Table 1.
Table. 1. PRM comparison with the best cutting-edge routing solutions
Cache Hit Ratio Average Query Delay Average Hop Count Power Consumption
Protocols Year
(CHR) (AQD) (AHC) (PC)

RSM 2022 97% 2.37 10 4720

RSM's primary benefit is the much reduced calculations and updating of routing tables. In contrast to
earlier algorithms, RSM produces the lowest AQD. RSM is hence an appropriate algorithm for real-time
systems.
5. Conclusion
This study presented a new Route Planning Algorithm (RPA).There are three key modules that make up
RPA: (i) Robot Finding Module (RFM), (ii) Robot Charging Module (RCM), and (iii) Route Selection
Module (RSM). In order to reduce the risk of viral propagation and perhaps save thousands of lives, it is
crucial to use such autonomous systems as RPA in settings where people must congregate. The
simulation results using the suggested framework demonstrate the flexible and effective mobility of the
robots under diverse environmental conditions as opposed to conventional methods. RSM is contrasted
with the leading cutting-edge topology routing options. RSM's primary benefit is the much reduced
calculations and updating of routing tables. In contrast to earlier algorithms, RSM produces the lowest
AQD. RSM is hence an appropriate algorithm for real-time systems.
References
[1] Wuhan lockdown 'unprecedented', shows commitment to contain virus: WHO representative in China -
Reuters. January 23, 2020.
[2] Zhang SX, Wang Y, Rauch A, Wei F. Health, distress, and life satisfaction of people one-month into COVID-
19 outbreak in China. medRxiv. 2020 Jan 1.

10
[3] Wang C, Horby PW, Hayden FG, Gao GF. A novel coronavirus outbreak of global health concern. The Lancet.
2020 Feb 15;395(10223):470-3.
[4] (WHO) Emergency Committee. Statement on the second meeting of the International Health Regulations
(2005) Emergency Committee regarding the outbreak of novel coronavirus (COVID-19). Geneva: WHO, 2020.
[7] World Health Organization (WHO). https://www.weforum.org/agenda/2020/04/10-april-who-briefing-health-
workers-covid-19-ppe-training/ retrieved on 10 April, 2020.
[8] Schwartz J, King CC, Yen MY. Protecting Health Care Workers during the COVID-19 Coronavirus Outbreak–
Lessons from Taiwan's SARS response. Clinical Infectious Diseases. 2020 Mar 12.
[9] World Health Organization. Coronavirus disease (COVID-19) outbreak: rights, roles and responsibilities of
health workers, including key considerations for occupational safety and health, 18 March 2020.
[10] https://www.nytimes.com/interactive/2020/03/13/world/asia/coronavirus-death-life.html.
[11] https://www.thedailybeast.com/covid-19-is-killing-italys-doctors-the-us-could-be-next.
[12] J. Zico Kolter, Andrew Y. Ng , “The Stanford LittleDog: A Learning and Rapid Replanning Approach to
Quadruped Locomotion”, June 21, (2010).
[13] Marco Hutter, et al., “ANYmal - A Highly Mobile and Dynamic Quadrupedal Robot”, (2016),
https://doi.org/10.1109/IROS.2016.7758092.
[14] Gerardo Bledt, “MIT Cheetah 3: Design and Control of a Robust, Dynamic Quadruped Robot”, Conference
Paper · October (2018), DOI: 10.1109/IROS.2018.8593885.
[15] Dan Xiang, Hanxi Lin, Jian Ouyang, Dan Huang, “Combined improved A* and greedy algorithm for path
planning of multi-objective mobile robot”, August 2022, Scientific Reports 12(1), DOI: 10.1038/s41598-022-
17684-0
[16] Talaat, F.M., Gamel, S.A. RL based hyper-parameters optimization algorithm (ROA) for convolutional neural
network. J Ambient Intell Human Comput (2022). https://doi.org/10.1007/s12652-022-03788-y

11

View publication stats