Joint Task Allocation and Path Planning for Space

Robot
Yifei Sun
Guangdong University of Technology
Zikai Zhang
Beijing Jiaotong University
YiDong Li
Beijing Jiaotong University
Jigang Wu (  [email protected] )
Guangdong University of Technology

Research Article

Keywords: Space robot, Path planning, Combination algorithm, Completion latency

Posted Date: July 19th, 2022

DOI: https://doi.org/10.21203/rs.3.rs-1844809/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
 Joint Task Allocation and Path Planning for Space
Robot

Yifei Suna,b , Zikai Zhangb,c , YiDong Lib , Jigang Wua,∗

a School of Computer Science and Technology, Guangdong University of Technology
b School of Computer and Information Technology, Beijing Jiaotong University
c State Key Lab of Rail Traffic Control and Safety, Beijing Jiaotong University

Abstract
Space robots have a wide application prospect in the aerospace industry. Due
to the limited fuel for the space robots, which cannot support the space robots
to run for a long time. In addition, the special working environment makes it
impossible to replenish the fuel for the robot at any time. Therefore, a proper
path is crucial for the operation of space robots. In order to ensure the op-
eration of space robots, we optimize the allocation of exploration tasks and
the selection of space robot paths, jointly. We propose two combination al-
gorithms named Parallel Search and Task Allocation (PS-TA) and Subbranch
Insertion and Task Allocation (SI-TA) to optimize the path and the task allo-
cation, intend to obtain the minimum completion latency. We also construct
Random Path Planning and Task Allocation (RTA) as the baseline. At last,
we provide extensive experiments to demonstrate that proposed algorithms can
obtain lower completion latency compared with RTA. Furthermore, SI-TA is
more energy-efficient than PS-TA.
Keywords: Space robot, Path planning, Combination algorithm, Completion
latency

1. Introduction
Space robot technology obtain continuous progress with the sustainably de-
veloped space industry. The space robot can perform various space tasks such
as area exploration, and sample collection without astronauts [1, 2]. Generally,
space robots can be categorized into on-orbit and exploration robots. The on-
orbit robots mainly provide diverse on-orbit services. While the main services
provided by exploration robots mainly include extravehicular exploration and
base construction, etc.
Advances in the space field have greatly promoted the research of space
robots, such as path planning of space robots and system design [3, 4]. As early

∗ Correspondingauthor
Email address: [email protected] (Jigang Wu)

Preprint submitted to Computers & Electrical Engineering July 11, 2022

as 1994, some scholars have already systematically studied the path planning
problem [5]. The path planning can be classified into local path planning and
global path planning, depending on the knowability of global information [6, 7].
Global path planning is implemented under the condition that global informa-
tion is known. The three most established global path planning methods are
intelligent bionic algorithm, graph search algorithm, and random sampling al-
gorithm [8]. In the static environment, global path planning can provide the
optimal path for each space robot [9]. While the local path planning (online
path planning) performs better in the dynamic environment [10]. Local path
planning can adaptively generate the current optimal trajectory when obstacles
occur between the source and destination [11]. To obtain optimal path planning,
many algorithms have been adopted such as the simulated annealing algorithm
[12], and fuzzy logic algorithm [13].
Since the practical robots have limited resources for space work, path plan-
ning is a crucial issue for studying to improve the efficiency of planet explo-
ration. Path planning is a typical complex nonlinear optimization problem.
Traditional optimization strategies are ineffective in resolving the issue. The
bionic has been widely adopted to solve complex optimization since its appear-
ance. Researchers have proposed different biological-based solutions to tackle
path planning for space robots [14, 15, 16, 17, 18, 19]. The above works mainly
study the single space robot path planning. Few works focus on multi-space
robot path planning.
In this paper, we mainly investigate the joint task allocation and path plan-
ning for multiple space robots. Each exploration region must be explored and
only require exploration once to decrease overall completion delay. Multiple
space robots set out simultaneously from different starting areas to perform ex-
ploration tasks and assemble in a single destination after all areas have been
explored. The most similar problem to this paper is the Multi-Depots Vehicle
Routing Problem (MD-VRP) [20]. In MD-VRP, multiple vehicles start from
different depots, serve a set of customers, and terminate their tours at the de-
pots. The time spent by a vehicle moving between two customers is much longer
than the time consumption spent serving customers. However, in this paper, the
moving time of a space robot is much less than the exploration time. Therefore,
the algorithms used for MD-VRP can not be applied to our problem directly.
Although an exploration area may appear in multiple paths, only one space
robot need to explore this area. This makes us aware of the importance of a
proper task allocation strategy. It is difficult to predict the time of a space
robot moving in two areas and exploring one area in advance. In this case,
space robots can only generate one-step trajectory after each actions, including
moving and exploring. It is vital to considerably allocate the exploration tasks
while optimizing path planning to lower the exploration latency [21].
To our knowledge, this is the first research to look into multi-space robot task
allocation and path planning, jointly. The main contributions are as follows.
• We investigate the joint exploration task allocation and path planning for
space robots. Multiple space robots locate their respective starting areas

2
 and set off at the same time to execute exploration tasks. Finally, all
space robots assemble at the same destination after completing tasks. We
model the problem to minimize exploration delay.
• We propose two combination algorithms named SI-TA and PS-TA. The
sequential search subbranch insertion in SI-TA allocates the remaining
areas according to the graph construction and initial path of each space
robot. The forward parallel search algorithm in PS-TA finds the next area
in each round depending on its distance to the destination. Based on the
paths generated by SI-TA and PS-TA, a Greedy-based Task Allocation
(GRTA) algorithm based on greedy policy is proposed to minimize the
exploration latency.
• We create the RTA algorithm as the baseline. Then, we implement exten-
sive experiments to demonstrate the effectiveness of SI-TA and PS-TA.
PS-TA achieves lower compared with SI-TA, while SI-TA consumes less
fuel than PS-TA, according to the results of the experiments. Moreover, in
terms of completion delay and fuel usage, PS-TA and SI-TA all outperform
RTA.
The remainder is organized as follows. We summarize the related work in
Section 2. We then describe the studied problem and model the problem in
Section 3. In Section 4, we propose two algorithms named PS-TA and ST-TA,
respectively. Extensive experiments and analyses are presented in Section 5
Finally, the conclusion of this paper is presented in 6.

2. Related Works

Path planning in dynamic situations is one of the most essential study issue
of robots industry, especially when the extreme conditions are subject to robots
in space environment.
Since the path planning problem is NP hardness [22], it is hard for a ex-
ponentially complicated traverse algorithm to satisfy the time limited practical
applications Path planning is also a fundamental question in mobile robots to
generate an optimal or sub-optimal path to enable the efficient operation of
the robot [23, 24, 25]. In general, path planning approaches may be divided
into three categories: biologically inspired, combinatorial, and sampling-based
[26]. The biological-based path planning is a type of intelligent algorithm that
simulates the evolutionary behaviors of biology. The combinatorial path plan-
ning usually integrates the graph search algorithm and workspace representation
methods to optimize the path planning problem [27]. The sampling-based algo-
rithms are usually utilized to find a path quickly [28]. However, the path found
by sampling-based algorithms may not be the optimal path.
In recent years, biologically-based techniques have been increasingly popular
in space robot path planning. The authors in [14] presented a genetic algorithm
to find the path to ensure that the on-orbit space robots get close to the tar-
get satellites, safely. The authors in [15] developed a self-adaptive ant colony

3
algorithm, which has a faster convergence rate than the traditional ant colony
algorithm. To avoid the obstacle, the authors in [16] proposed an artificial bee
colony algorithm. In addition, some academics have developed a series of combi-
nation techniques to boost path planning performance even more. The authors
in [17] integrated a genetic algorithm and simulated an annealing algorithm to
generate a path with the capability of obstacle avoidance. In a 3D environment,
a hybrid genetic-cuckoo search algorithm was designed in [18] to generate an
optimal path and can alleviate the conflict between optimality and delay.
Except for the biology-based algorithms, the authors of [19] suggested a path
planning algorithm on the basis of enhanced fuzzy control to avoid the obstacles
that appeared continuously. The authors of [29] designed a deep deterministic
policy gradient method with multi-constrained reward to find the best path
while keeping path length, coupling disturbance, and safety in mind.
The existing works mentioned above can solve the path planning problem
for single space robots with different methods. However, the cooperation of
multiple space robots in a space environment is less considered to the most of
our knowledge.

3. Scenario description and problem definition

In our studied scenario, multiple space robots are required to execute multi-
ple areas exploring tasks, collaboratively. To avoid additional fuel consumption,
we assume that each area needs to be explored only once. All space robots un-
dertake exploration activities at the same time in different starting locations and
finally stay in the same destination area when all areas have been explored. In
our work, we conjointly study the execution decisions of exploration tasks and
the multi-space robot path planning to minimize total exploration task com-
pletion delay. Besides, we consider homogeneous space robots for simplicity. A
space robot does not take time to explore the area which has been explored by
other space robots. Note that, a space robot is required to explore the area
which is not explored as it arrive. If some space robots arrive in a previously
unexplored area at the same time, the lower serial number space robot explore
this area. We will cover the serial number later.
Next, we will introduce the studied scenario and problem in this paper and
then formulate the problem. We define an undirected graph G = (V, E) to
abstractly represent the entire exploration area. We denote E = {1, 2, · · · , e}
to represent the paths between areas. Besides, let V = {1, 2, · · · , v} be the
area set. In order to facilitate the description, we will use word “node” and
“area” interchangeably in the following text, indicating the same meaning. For
an exploration task, we can obtain the corresponding graph G = (V, E) in
advance. Then, we number the nodes in a graph. We set the serial number of
the starting node to start at 0, that is 0, 1, · · · , s − 1. Space robots at different
starting nodes have the same serial number as the starting nodes where they
locate, which also means that there are s space robots. Note that the destination
is the node with the largest serial number. Let λi,j b indicate whether the space

4
robot b moves between area i and j.
(
i,j 1 The space robot b moves between area i and j
λb = (1)
0 otherwise.

When i = j, λi,j b = 0. We can use an order sequence h·i to represent a space

robot path. For example, if a space robot locate at area 1 in the begin, goes
through area 2 and comes back to area 1, then the path of this space robot
can be represent as h1, 2, 1i. Next, a discriminant operation ⊖ was defined as
follows. If λi,j i,j i,j i,j
b = 0, λb ⊖ j = {∅}. When λb = 1, λb ⊖ j = {j}. Then, a
splicing operation ⊗ between sets is also defined. The rule for ⊗ is as follows.

{f } ⊗ {g} = hf, gi. (2)

F
Finally, let be the operation of set accumulation.
v
G
{φa } = {φa } ⊗ {φa+1 } ⊗ · · · ⊗ {φv }. (3)
a

Then, let Pb be the path of space robot b, which can be obtained by follow
formula.
v−1
G v−1
G i,j
Pb = {b} ⊗ ( λb ⊖ j), (4)
i=b j=0

Our main goal is to achieve minimum exploration latency. We define the time
for space robot b to explore area i as τbi . If the multiple paths contain the
same area, then, which space robot takes the responsibility for the exploration
of this area should be determined carefully. Let γbi indicate whether the space
robot b performs the exploration tasks of area i, where b ∈ {0, 1, · · · , s − 1},
i ∈ {0, 1, · · · , v − 1}.
(
i 1 The space robot b explore area i
γb = (5)
0 otherwise.

Let Tbexe denote the time spent by space robot b completing the exploration
task.
v−1
X
Tbexe = γbi · τbi . (6)
i=0

Let κi,j
b indicate the delay for space robot b moving between area i and area j.
Then, we denote θ to indicate the position of a node in Pb , where 1 ≤ θ < |Pb |.
As an illustration, the θ-th node on path Pb can be represented as Pb [θ]. Let
Tbr be the overall movement delay of space robot b.
|Pb |−1
P [θ],Pb [θ+1]
X
Tbr = κb b . (7)
θ=1

5
We denote T to represent as the overall task completion latency.

T = max{Tbexe + Tbr }. (8)

We can formalize the latency minimization problem as follows.

min T (9)
s−1 v−1
X X
s.t. γbi = v. (10)
b=0 i=0

Formula (10) ensures that each area is explored.

Algorithm 1 Find Shortest Path

Input: s, v, G
Output: l
1: Initial l
2: Append s to l //Append start node to path l
3: while l does not end with v do
4: Get the last node e of l
5: Obtain the neighbor node set C of e according to G
6: Calculate the distance d between first node in set C and v
7: de,v = Caldis(e, v) //Calculate the distance between two nodes
8: for each c in C do dc,v = Caldis(c, v)
9: if dc,v<de,v and dc,v < d then
10: d ← dc,v
11: Obtain the node c∗ corresponding to d
12: Append c∗ to l
13: return l

4. Solutions

The joint optimization problem is decomposed into two subproblems, path

planning, and task allocation. Before solving the task allocation problem, we
need to find the working path for all space robots. The path of each space records
the movement sequence of the robot from a starting area to the destination. The
generated paths must cover all nodes in the given graph. Then, we solve the
task allocation problem to obtain the execution decision of exploration tasks
to minimize the overall latency. In particular, we propose two path planning
algorithms named Sequential Search Subbranch Insertion (SSSI) and Forward
Parallel Search (FPS), respectively. After that, a task allocation algorithm
named Greedy-based Task Assignment (GRTA) is proposed to generate the
execution decision of each exploration area. Finally, we combine SSSI with
GRTA to obtain SI-TA and combine FPS with GRTA to obtain PS-TA.

6
4.1. Solutions for Path Planning
We next introduce the SSSI and FPS, respectively. We first introduce SSSI.
The Algorithm 1 is used to obtain the shortest path for each space robot, which
is usually used to generate the shortest path in many works. However, the
paths obtained by using Algorithm 1 can not cover all nodes. Next, the SSSI
algorithm is used to insert nodes that are not covered into the shortest paths
obtained by Algorithm 1. The details of SSSI are described as follows.

Algorithm 2 Sequential Search Subbranch Insertion

Input: G, lb , b ∈ {1, 2, · · · , s} //lb is initial path of space robots b
Output: lb
1: w = 0
2: Initialize Fb
3: Denote z to indicate state of nodes //explored or unexplored
4: while There exists paths b where lb [w] is not destination do
5: Φ = {∅}
6: Record serial number of all space robots that meet the conditions in Φ.
7: for each ϕ in Φ do
8: Construct hϕ //neighbor nodes set of lϕ [w]
9: if There exits nodes that not explored then
10: Insert the node lϕ [w] into the last position of Fϕ
11: Choose an unexplored node h′ in hϕ
12: Insert h′ into position lϕ [w + 1]
13: Update z
14: if The successor node g of lϕ [w] are contained by hϕ then
15: if The last node of Fϕ is in g then
16: Insert g into position lϕ [w + 1]
17: Remove g from Fϕ
18: w = w + 1;
19: return lb

First, stores all paths that do not reach the destination according to w,
where w is the search position on paths. Define set Φ to record the space robot
serial number corresponding to the paths. Then, we can construct the set hϕ to
record serial numbers of neighbor nodes of node lϕ [w] according to the structure
of the given graph G, where lϕ represents the path of robot ϕ.
Next, we judge the type of nodes in Φ according to z, where z is used to
indicate whether a node is explored. If the node lϕ [w] is an unexplored node,
we insert lϕ [w] into the last position of Fϕ , where Fϕ is the set of bifurcation
nodes. We denote a node as a bifurcation node if the neighbor nodes of this
node are not all explored nodes. Then, we select an unexplored node h′ in hϕ
and insert h′ into the position lϕ [w + 1]. We next update the z. If all nodes in
hϕ are explored. We insert node g into position lϕ [w + 1], where g is contained
by hϕ and is the successor node of lϕ [w] and the last node of Fϕ . Next, we
remove g from Fϕ . Finally, update the search position w. The pseudocode of
SSSI is described in Algorithm 2.

7
Algorithm 3 Distance Calculation
Input: G = {N , E}
Output: D //distance information
1: µ = 1
2: Initial E //indicator vector
3: while Sum(E) < |N | do //Sum(E) represent the sum of elements in E
4: Initialize ς //set of nodes that are not operated
5: Initialize η
6: for α = 0 to |op| − 1 do
7: for β = 0to |N | − 1 do
8: if op[α] is a neighbor of β then
9: if ind[β] == 0 then
10: D[β] = µ //record distance
11: E[β] = 1
12: append the node β to ς
13: µ=µ+1
14: η=ς
15: return D

Next, we introduce FPS. The main idea of FPS is to synchronously search

the next node to go for each space robot that has not reached the destination.
The successful execution of FPS requires preliminary knowledge of the distance
from all nodes to the destination. We adopt Algorithm 3 to obtain the distance
information. In this process, we assume that the distance between two adjacent
nodes is 1. We first set the destination node to be the initial node. Then,
the distance of nodes that are directly adjacent to the destination and have
not been accessed is set as 1. Continue to find the nodes which are directly
adjacent to the nodes with 1 distance to the destination and have not been
accessed, and the distance of these nodes to the destination is set as 2. Repeat
the above steps until all nodes have been accessed. We denoted D to record
the distance information. After that, we obtain the path of each space robot by
using Algorithm 4. The main process is summarized as follows.
Step1: Find all space robots which have not reached the destination and con-
struct set S to store the found space robots serial number.
Step2: Construct set Cr of neighbor nodes of the last node for each pr accord-
ing to the given G. Each node in Cr may be an explored node or an
unexplored node. We denoted Wr to represent the set of unexplored
nodes.
Step3: Calculate the number ηr of nodes in Cr .
Step4: Append the only node in Wr to lr when ηr = 1, where lr is the path of
space robot r.
Step5: Find out all nodes which are farthest to the destination according to
r
D and Wr . Construct Dmax to record the corresponding space robots
serial numbers.

8
Algorithm 4 Forward Parallel Search
Input: G = {N , E}, D
Output: li
1: Initialize ρ
2: Initialize li to record path for i
3: while There are robot r ∈ S that has not reached destination node do
4: for each r in S do
5: Get neighbor nodes set Cr of last node of path according to the G
6: Get unexplored nodes set Wr of i according to Cr and ρ
7: Calculate the number ηr of nodes in Wr
8: if ηr == 1 then
9: Append this node to path lr
10: Update ρ
11: if ηr > 1 then
r
12: Construct set Dmax of nodes that are farthest to destination
node according to D, Wr
r
13: if Dmax only contains one node then
14: Append this node to path lr
15: else
r
16: Select a node from Dmax randomly to append to lr
17: Append the penultimate node on lr to lr again
18: Update ρ
19: if ηr ≤ 0 then
20: while true do
21: Find node with minimal distance to destination node in Cr
22: Append this node to lr
23: if r has reached the destination node then
24: break //terminates the path search for r
25: Repeat line 5 to 18
26: break;
27: return li

9
 r
Step6: Choose a node from Dmax to add to the last position of lr . After that,
we append the penultimate node on lr to lr again.
Step7: Find out the node with minimal distance from starting node to destina-
tion in Cr and append this node to lr . If r reaches the destination, we
terminate the algorithm. Otherwise, repeat Step4 - Step6 once. Then,
we terminate the algorithm.
For the pseudocode of FPS, see Algorithm 4

4.2. Greedy-based Task Assignment

In this subsection, we propose GRTA based on greedy policy to generate the
execution decisions of exploration tasks. The details of GRTA are described in
Algorithm 5.

Algorithm 5 Greedy-based Task Allocation

Input: pb
Output: T
1: w = 1, γ = pb [0]
γ
2: T = τb
3: Let vector z indicate whether nodes are explored
4: Let mb represent the length of pb
5: while There exists w ≤ mb do
6: U = {∅}
7: for b = 0 to s − 1 do
8: if w ≤ mb then
9: U =U ∪b
10: for each u in U do
11: if Node pu [w] is an explored node then
p [w−1],pu [w]
12: Tu = Tu + κuu
13: else
14: Construct set Ω to record paths whose last node is pu [w]
15: Select space robot a whose Ta is minimum.
p [w] p [w−1],pa [w]
16: Ta = Ta + τa a + κaa
17: for Robot α in Ω \ a do
p [w−1],pα [w]
18: Tα = Tα + καα
19: Update z
20: w =w+1
21: Compute T with formula (8)
22: return T

First, let mb be the length of path pb and w be the index to indicate the
current search position of all paths. We define U to record the space robot serial
number. If w ≤ mb , we append b to U . Then, we judge whether the node pb [w]
has been explored for each space robot b. If node pb [w] has been explored, we
just compute the moving delay for b arriving at this node. If this node is an

10
   
  


   
   


    

     

  
       
      

    
 
(a) Graph G1 (b) Graph G2

Figure 1: Graphs for experiment

unexplored node, we record all paths whose node in position w is this node.
Then, we construct Ω to store the robot serial number corresponding to these
paths. We next choose a node a from Ω with a minimum Ta and compute the
exploring time and moving time of a. For node α in Ω \ a, we only compute the
movement time of the robot arriving at this node. Finally, we update the state
of nodes in ρ. We calculate the T with Equation (8) when h > max(Si ).
GAT can be formally described by Algorithm 5.

4.3. Combination algorithm

We combine Algorithm 2 and Algorithm 5 to obtain SI-TA algorithm. Be-
sides, we also obtain a combination algorithm named PS-TA by combining Al-
gorithm 4 and Algorithm 5. SI-TA and PS-TA solve the path planning problem
through Algorithm 2 and Algorithm 4, respectively. After that, SI-TA and PS-
TA obtain the execution decision of exploration task by using GRTA algorithm.

5. Simulation Results and Analysis

A series of experiments are implemented in this section to demonstrate the

performance of SI-TA and PS-TA. Furthermore, RTA algorithm is created for
comparison. The RTA selects nodes randomly for each space robot and uses
GRTA to make decisions for exploration tasks.

5.1. Simulation Setup

In the experiments, we construct a specific scenario where three space robots
execute an exploration task. As shown in Figure 1(a)[30] and Figure 1(b) which
are graphs abstracted from two areas. The first exploration task consists of 20
exploration areas and the second exploration task consists of 15 exploration
areas. In the beginning, all space robots stay at their respective starting areas
(0, 1, and 2). We set the largest serial number node as the destination. We set
the moving delay between two areas is uniformly distributed in [10, 20]s. The
time consumed by exploring an area is distributed in [200 500]s.

11
 2
PS-TA
1.8 SI-TA
RTA
1.6

1.4
Normalized latency

1.2

0.8

0.6

0.4

0.2

0
Experimental graph G 1 Experimental graph G 2

Figure 2: Completion latency.

2
SI-TA
1.8 PS-TA
RTA

1.6
Normalized moving time

1.4

1.2

0.8

0.6

0.4

0.2

0
Experimental graph G 1 Experimental graph G 2

Figure 3: The movement delay obtained by algorithms.

12
 0.03 0.02
PS-TA PS-TA
SI-TA 0.018 SI-TA
0.025
0.016

0.014
0.02
Runnding time

Runnding time
0.012

0.015 0.01

0.008
0.01
0.006

0.004
0.005
0.002

0 0
0 1 2 3 4 5 6 7 8 9 10 11 0 1 2 3 4 5 6 7 8 9 10 11
Random execution Random execution

(a) On graph G1 (b) On graph G2

Figure 4: Algorithms running time on two graphs

5.2. Experimental Results and Performance Analysis

First, we focus on the performance of algorithms on task completion latency.
We normalize the latency by dividing the same constant. The Figure 2 shows
the final results. It is observed that the latency of PS-TA is minimum. Because
PS-TA can make the load between space robots more balanced. We can also
see that the delay of RTA is the largest, mainly due to the high randomness of
the path planning algorithm used by RTA.
Because space robot fuel is limited, and refueling is difficult. Therefore,
it is significant to consider the fuel consumption of space robots. Due to the
homogeneity of space robots, the exploration time for all space robots to explore
an area is same under different algorithms. The main difference in completion
latency of different algorithms main depends on the moving delay. Therefore, we
can compare the energy efficiency of algorithms according to the space robots
moving delay. For the sake of simplicity, we normalized the movement time.
The Figure 3 show the experimental results. From the figure, we can see that
the SI-TA can save more fuel than PS-TA. The fuel consumption of RTA is the
highest. This is caused by the randomness of path planning in RTA.
Next, the running time of the SSSI and the FPS are counted. To avoid
contingency, we count the results of 10 experiments, respectively. The Figure
4(a) and Figure 4(b) show the results on G1 a nd G2 , respectively. From Figure
4, the running time of SSSI is higher. The main reason is that SSSI spent more
time executing the insert operation.

6. Conclusion

In this paper, we have investigated the joint path planning and task allo-
cation problem for space robots and formalized the problem to minimize the
exploration task completion latency. In this problem, multiple space robots set
out simultaneously from different starting areas to perform exploration tasks

13
and assemble at a single destination after all areas have been explored. We
have proposed two algorithms named SI-TA and PS-TA by integrating a task
allocation algorithm and a path planning algorithm to solve the completion la-
tency minimization problem. In particular, we have designed two path planning
algorithms named SSSI and FPS, which can generate paths for all space robots
moving from their respective starting node to the destination. In addition, each
node of the given graph must be contained by at least one path. Furthermore,
we have proposed an algorithm named GRTA as the task allocation algorithm
to allocate the exploration tasks based on the generated paths to space robots
with the greedy idea to obtain minimum completion latency. To further prove
the performance of the proposed algorithms, the RTA algorithm has been con-
structed as a baseline, which generates paths for space robots by randomly
selecting nodes and allocating tasks with GRTA. To demonstrate the perfor-
mance of proposed algorithms, extensive experiments have been implemented.
Experimental results show the effectiveness of the proposed algorithm. In terms
of fuel consumption and completion latency, the proposed algorithms perform
better compared with RTA. The proposed algorithms are all better than RTA
on delay and fuel consumption. Specifically, the PS-TA can save more time to
complete the exploration tasks than SI-TA. Besides, SI-TA consumes less fuel
than PS-TA.

7. Acknowledgement

Funding:This work is supported by the National Natural Science Foundation

of China [No. 62106052 and 62072118], Guangdong Key R&D Project of China
[No. 2019B010121001].

8. Data availability

The data underlying this article will be shared on reasonable request to the
corresponding author.

9. Declarations

Conflict of interest The authors declare that they have no conflict of

interest.

References

[1] Z. XING, Y. ZHAO, S. Zhu, Path planning method design and dynamic
model simplification of free-flying space robot, in: 2020 15th IEEE Confer-
ence on Industrial Electronics and Applications (ICIEA), 2020, pp. 1501–
1505. doi:10.1109/ICIEA48937.2020.9248286.
[2] Y. Lin, D. Li, Y. Wang, et al., Current status and analysis of space robot,
Spacecr. Eng 24 (5) (2015).

14
 [3] L. E. Xiong Youlun, Ding Han, Robotics, Mechanical Industry Press, Bei-
jing, 1993.
[4] A. N. Atiyah, N. Adzhar, N. I. Jaini, An overview: on path planning op-
timization criteria and mobile robot navigation, Journal of Physics: Con-
ference Series 1988 (1) (2021) 012036. doi:10.1088/1742-6596/1988/1/
012036.
[5] L. Chen, Y. Huang, H. Zheng, H. Hopman, R. Negenborn, Cooperative
multi-vessel systems in urban waterway networks, IEEE Transactions on
Intelligent Transportation Systems 21 (8) (2020) 3294–3307. doi:10.1109/
TITS.2019.2925536.

[6] P. B. Kumar, C. Sahu, D. Parhi, K. Pandey, A. Chhotray, Static and dy-

namic path planning of humanoids using an advanced regression controller,
Scientia Iranica 26 (1) (2019) 375–393.
[7] Z. Tang, H. Ma, An overview of path planning algorithms, IOP Conference
Series: Earth and Environmental Science 804 (2) (2021) 022024. doi:
10.1088/1755-1315/804/2/022024.
[8] K. Cai, C. Wang, J. Cheng, C. W. De Silva, M. Q.-H. Meng, Mobile
robot path planning in dynamic environments: a survey, arXiv preprint
arXiv:2006.14195 (2020). doi:10.15878/j.cnki.instrumentation.
2019.02.010.

[9] G. Klancar, A. Zdesar, S. Blazic, I. Skrjanc, Wheeled mobile robotics: from

fundamentals towards autonomous systems, Butterworth-Heinemann.
[10] M. M. Costa, M. F. Silva, A survey on path planning algorithms for mobile
robots, in: 2019 IEEE International Conference on Autonomous Robot Sys-
tems and Competitions (ICARSC), 2019, pp. 1–7. doi:10.1109/ICARSC.
2019.8733623.
[11] H. S. Dewang, P. K. Mohanty, S. Kundu, A robust path planning for mobile
robot using smart particle swarm optimization, Procedia Computer Science
133 (2018) 290–297, international Conference on Robotics and Smart Man-
ufacturing (RoSMa2018). doi:10.1016/j.procs.2018.07.036.
[12] B. Basbous, 2d uav path planning with radar threatening areas using sim-
ulated annealing algorithm for event detection, in: 2018 International Con-
ference on Artificial Intelligence and Data Processing (IDAP), 2018, pp.
1–7. doi:10.1109/IDAP.2018.8620881.

[13] A. Bakdi, A. Hentout, H. Boutami, A. Maoudj, O. Hachour, B. Bouzouia,

Optimal path planning and execution for mobile robots using genetic algo-
rithm and adaptive fuzzy-logic control, Robotics and Autonomous Systems
89 (2017) 95–109. doi:10.1016/j.robot.2016.12.008.

15
[14] A. Seddaoui, C. M. Saaj, Collision-free optimal trajectory for a controlled
floating space robot, in: Towards Autonomous Robotic Systems, Springer
International Publishing, Cham, 2019, pp. 248–260.
[15] W. Ye, D. Ma, H. Fan, Path planning for space robot based on the self-
adaptive ant colony algorithm, in: 2006 1st International Symposium on
Systems and Control in Aerospace and Astronautics, 2006, pp. 4 pp.–33.
doi:10.1109/ISSCAA.2006.1627696.
[16] F. Jin, G. Shu, Path planning of free-flying space robot based on artificial
bee colony algorithm, in: Proceedings of 2012 2nd International Conference
on Computer Science and Network Technology, 2012, pp. 505–508. doi:
10.1109/ICCSNT.2012.6525987.
[17] F. Jin, Path planning of free-flying space robot using memetic algorithm,
in: Ifost, Vol. 2, 2013, pp. 19–22. doi:10.1109/IFOST.2013.6616889.
[18] J. Wang, X. Shang, T. Guo, J. Zhou, S. Jia, C. Wang, Optimal path
planning based on hybrid genetic-cuckoo search algorithm, in: 2019 6th
International Conference on Systems and Informatics (ICSAI), 2019, pp.
165–169. doi:10.1109/ICSAI48974.2019.9010519.
[19] B. Shi, H. Wu, Space robot motion path planning based on fuzzy control
algorithm, Journal of Intelligent and Fuzzy Systems (Preprint) 1–8.
[20] Y. E. M. Vieira, R. A. de Mello Bandeira, O. S. da Silva Júnior, Multi-depot
vehicle routing problem for large scale disaster relief in drought scenarios:
The case of the brazilian northeast region, International Journal of Disaster
Risk Reduction 58 (2021) 102193.
[21] Y. Sun, Z. Zhang, Y. Li, J. Wu, Joint optimization of path planning and
task assignment for space robot, in: 2021 12th International Symposium
on Parallel Architectures, Algorithms and Programming (PAAP), IEEE,
2021, pp. 47–51.
[22] B. Chen, G. Quan, Np-hard problems of learning from examples, in: 2008
Fifth International Conference on Fuzzy Systems and Knowledge Discovery,
Vol. 2, 2008, pp. 182–186. doi:10.1109/FSKD.2008.406.
[23] K. Karur, N. Sharma, C. Dharmatti, J. E. Siegel, A survey of path planning
algorithms for mobile robots, Vehicles 3 (3) (2021) 448–468.
[24] Y. Zhuang, Y. Sun, W. Wang, Mobile robot hybrid path planning in an
obstacle-cluttered environment based on steering control and improved dis-
tance propagating, Int. J. Innov. Comput. Inf. Control 8 (2012) 4095–4109.
[25] M. A. Contreras-Cruz, V. Ayala-Ramirez, U. H. Hernandez-Belmonte, Mo-
bile robot path planning using artificial bee colony and evolutionary pro-
gramming, Applied Soft Computing 30 (2015) 319–328. doi:10.1016/j.
asoc.2015.01.067.

16
[26] S. K. Debnath, R. Omar, N. B. A. Latip, A review on energy efficient
path planning algorithms for unmanned air vehicles, in: R. Alfred, Y. Lim,
A. A. A. Ibrahim, P. Anthony (Eds.), Computational Science and Technol-
ogy, Springer Singapore, Singapore, 2019, pp. 523–532.
[27] S. M. LaValle, Planning algorithms, Cambridge university press, 2006.

[28] F. Yan, E. Xia, Z. Li, Z. Zhou, Sampling-based path planning for high-
quality aerial 3d reconstruction of urban scenes, Remote Sensing 13 (5)
(2021) 989.
[29] X. Hu, X. Huang, T. Hu, Z. Shi, J. Hui, Mrddpg algorithms for path
planning of free-floating space robot, in: 2018 IEEE 9th International Con-
ference on Software Engineering and Service Science (ICSESS), 2018, pp.
1079–1082. doi:10.1109/ICSESS.2018.8663748.
[30] Mastermind – 2021 national algorithm design challenge, https://www.
shenjims.com/.

17