PATH PLANNING ALGORITHM UNDER SPECIFIC CONSTRAINTS

IN WEIGHTED DIRECTED GRAPH

Qiyun Sun, Wanggen Wan, Guoliang Chen, Xiang Feng

School of Communication and Information Engineering

Institute of Smart City, Shanghai University
Shanghai, China
[email protected], [email protected], [email protected], [email protected]

ABSTRACT is a kind of heuristic search algorithm which is based on

Dijkstra algorithm, and it is a best-first search algorithm.
At present, most of the path planning algorithms are only The main feature of the A* algorithm is that the known
aimed at reaching the end point from the starting point. road network information is introduced, especially about
However, in practical applications, the existence of a the target point information, in the choice of the next node
variety of constraints increases the difficulty of path of the path. Some objective function(such as the shortest
planning , so the Dijkstra algorithm, A* algorithm and distance, the minimum time etc.) is calculated between
other classical algorithms become no longer applicable in every candidate node and termination node, which is used
such a situation. In this paper, considering the various as evaluation indicator of whether a candidate node is the
constraints in practical problems, we abstracted out such best path node or not, and the candidate node with the
kind of path planning problem: In a weighted directed preferred objective function will be selected as the next
graph, paths need to be found which start from a source path node. Tan BaoCheng, Wang Pei[10] has improved the
node, after passing through some specified intermediate A* algorithm with the result that the path length and
nodes without repetition, and finally ends at a specified running time of algorithm was significantly better than the
termination node. Based on Dijkstra algorithm and A* algorithm. J. Yao, C. Lin et al.[11] introduced the
heuristic search principle, a feasible path planning evaluation function to simplify the search steps of A*
algorithm is proposed with the new concept of " inspired algorithm and solved the virtual human’s obstacle
hop count " to balance the path weight and search time. avoidance and navigation issues. In addition to the
traditional algorithms such as Dijkstra and A*, algorithms
Index Terms— path planning, Dijkstra, weighted combined with the biology are also emerging. A multi
directed graph, heuristic search,specified node set model and multi standard path planning algorithm based on
genetic algorithm was presented in the paper[12].
1. INTRODUCTION Paper[13] proposed a path planning method based on ant
colony algorithm. However, in graph theory, there are still
Path planning is the focus of many fields of technology such a class of problems which have not been extensively
research, which has broad application prospects and studied but have important practical significance:
practical value.Applications in high-tech fields are:UAV 1) "Exhibition Visitors problem". As for visitors in
flight [1],USV route planning [2],automatic vehicle driving large-scale exhibition, the problem is how to design a
[3],robot automation[4-5] and so on.Applications in the visiting route which makes visitors start from somewhere,
field of daily life are: automatic parking [6], road planning after visiting some pre-planned booths, arrived at
based on the GIS system, etc. destination with the total path shortest.
The core of path planning technology is its algorithm 2) "Traveler problem". To design a non repeating
design. From the traditional algorithm to the subsequent itinerary, considering traffic situation, so the traveler
algorithm with combination of biology, intelligent traveled througth all the pre-planned tourist attractions with
algorithms have made great progress. Many domestic and the total travel time shortest.
foreign scholars have conducted extensive research, and 3) "Bus route design problem". To design a bus line,
put forward a lot of schemes and algorithms for path so that the bus start from the start stop, after passing
planning. Dijkstra EW.A proposed famous Dijkstra through all the necessary stops without repetition, to reach
algorithm in 1959, this is a classic single-source shortest the terminal stop with the shortest travel distance.
path algorithm for calculating the shortest path from a node In addition, there are some special network routing
to all other nodes. And many people have ever done problems, such as the requirements of passing through
research on the shortest path problem based on Dijkstra certain intermediate routers. In order to solve the path
algorithm. J.R.Jiang, H.W.Huang et al. took the edge and planning problem with additional constraints, we need to
node weights into consideration and proposed an extended improve and innovate the original traditional path planning
Dijkstra algorithm [7]; Mingjun Wei and Yu Meng solved algorithm, and propose a new algorithm to solve these more
the problem about the optimal path of urban using complex problems. For the above such problems, we can
improved Dijkstra algorithm [8].A* (A-Star) algorithm [9] summarize as follows:

978-1-5090-0654-0/16/$31.00 ©2016 IEEE 635 ICALIP 2016

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.
 In a directed weighted graph, we need to consider the required to pass through the specified intermediate node set
weighted length of the path which starts from a specified {2,3} without repetition, then there are two paths:0→2→3
node, after passing through some specified intermediate →1 and 0→3→2→1,whose LinkID are 1|5|4 and 2|6|3
nodes without repetition (the number of intermediate nodes respectively,with corresponding weighted value 4 and 5.So
is uncertain), and finally ends at the specified termination the shortest path P is obtained as 0→2→3→1, it is an
node.And we aim to find such a feasible path hoping that acyclic directional path.
its weighted length would be as small as possible and
search time would be as short as possible. 2.1. Concept Definition
Most of the current studies of the shortest path
algorithm are just aimed at reaching the end point from the The following are definitions of some related terms:
starting point. Although some researchers considered path LinkID: Index of directed edge;
planning problem with constraints, such as the constraint- Cost: Weight of directed edge;
based multi-robot path planning algorithm given by the SourceID: The index of the starting node of the
M.Ryan[14]; Huang Shuli, Hu Dasha et al.[15] have directed edge;
studied about the shortest path algorithm in the unweighted DestinationID: The index of the terminate node of the
undirected graph with the constraint that the path must directed edgs;
cover a specified intermediate node set. But very few IncludingSet: Node set which must be passed
researchers have considered the complex path planning through, and different node indexes are split with the
problem with various constraints in a weighted directed Separator '|'.
graph as mentioned in the last paragraph. Weight of path: The sum of the weights of all the
For path planning issues raised above, due to the directed edges of the path.
additional constraints, it can not be solved by directly using Acyclic directed path P: A directed path from s
Dijkstra algorithm. Therefore, this paper takes the Dijkstra (starting node) to t (termination node) which is composed
algorithm as the foundation, and draws on the heuristic of a series of directed edges, and is not allowed to
search theory of A* algorithm, and finally gives a feasible repeatedly pass through any of the nodes; The
algorithm consulting the area routing calculation process representation of P is to list the index of the directed edges
of the OSPF routing protocol. In this algorithm, we put constituting the path in turn, and the indexes are split with
forward a new concept of "inspired hop count" ,it is the the Separator '|'.
most important parameter in this algorithm to dynamically
adjust the balance between the weight of path and the time 2.2. Problem Definition
of path searching. The experimental results show that this
algorithm can effectively solve the above path planning Given a weighted directed graph G = (V, E), V for the node
problem.
set, E for the directed edge set, each edge has its own
weight. For a given node s, t, a node set V and its subset ,
2. PROBLEM DESCRIPTION
to find a acyclic directed path P such that the path P can
travel through all nodes in without repitition(no limit to
Here first with a simple example of the problem
node traversal order in ). Note: All weights are integers
description. As shown in Figure 1, assuming the source
within [1,30]; Edgs connecting node A and B are likely to
node number is 0; the intermediate nodes need to pass
be more than one, whose weight may be the same or may
through are 2,3; And the end node number is 1.
be different.

0 3. PATH PLANNING ALGORITHM

Being different with general path planning problem, with

0,1 considering passing through a specified intermediate node
set in the path, the target node is no longer fixed and single,
1,2 2,1 but constantly changing, because each node in the
1 intermediate node set is required to be traveled before
reaching the final termination node, so every node in the
3,3 4,1 intermediate node set would be a temporary target node in
the process of searching path. This brings a lot of
5,1 uncertainty in the path planning, making it more complex.
2 6,1 3 LinkID,Cost However, the link-state algorithm for routing computation
Figure 1. Sample graph,a simple weighted directed graph
and the heuristic search theory of A* algorithm provide
inspiration and enlightenment for solving this complex path
planning problem. Therefore, in order to make readers
As to find out the shortest path from the source node
better understand the algorithm to be given, the paper first
0 to the end node 1, we can aquire the shortest path 0→1 introduces the process of routing computation of OSPF
according to the Dijkstra algorithm, whose LinkID is 0, routing protocol and the heuristic search theory of A*
with the weighted value 1;But if the path is additionally algorithm.

636

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.
 OSPF is an interior gateway protocols, based on link- nodes there are in the intermediate node set,the more
state algorithms.The process of routing computation by complex the problem will be), and then uses exhaustive
link-state algorithm is shown in Figure 2. RTA, RTB, RTC, search to list all the partial paths and selects one with
RTD are routers in a same area running OSPF, each router minimum path cost, making the chances of getting the
in the area floods Router-LSA (router link-state optimal path as large as possible. The specific algorithm
adervertisement) to form a LSDB(link state database) of the flow chart is shown in Figure 3.
area. Each router uses the SPF algorithm to calculate a
shortest path tree with itself as the root based on the LSDB Start
and this shortest path tree describes the shortest path to all
other routers in the area.The SPF algorithm used here is Load data: s, t, V', V
Dijkstra algorithm. Therefore, learn from the above process
of routing computation, the first step of our algorithm to be
Set inspired hop count: N
given is to calculate the shortest path tree of the nodes in
and the starting node according to the given graph by
using Dijkstra algorithm. To get SPT of nodes in
V' and source node s

LSDB
Is there a node in V' Y
RTA RTB LSA of RTA No solution
unreachable？
RTC LSA of RTB
Flooding
Router-LSA N
LSA of RTC
Calculate the cost of all
RTD LSA of RTD paths with top N hops

Back to the last hop and

take suboptimal node
SPF Algorithm
to recalculate

Remove the paths with a

repeating node
Take the current node
as the source node

N
Computing Path exists?
Routes
IP Routing Table Y
SPT
Figure 2. Routing computation of OSPF Take the first -hop nodes
of the shortest path
Not only the distance to the source node, but also the as the next -hop nodes
distance to the termination node of each node are taken into
consideration in A* algorithm. If g(n) is used to represent N Will reach
the path cost from the starting node to any other node n, the the termination node after N
h(n) represents the estimated path cost(heuristic value) of hops?
the node n to the termination node,define that:
Y
f(n) = g(n) + h(n)
A* algorithm calculates each node’s value of f(n) to find N=1? N=N-1
a minimum one in the path searching before reaching the Y
termination node. So, learn from the above process, our
algorithm to be given will also calculate each node’s value End
of the partial path cost based on the SPT acquired in the Figure 3. Flow chart of the proposed algorithm
first step to find the minimum one as the next node.
Data to be loaded includes a source node s (assume
3.1. Algorithm Description node number is 0), the termination node t (assume node
The algorithm to solve the path planning problem with a number is T-1), an intermediate node set to be passed
specified intermediate node set is mainly divided into two (assume that there are total S nodes whose node number are
steps. The first step is to focus on the source node s and the 1,2, ... S respectively, ie, = {1,2, ... S}), all node set
nodes in the intermediate node set, using the Dijkstra (assume that there are totally T nodes); Assuming that the
algorithm to calculate the shortest path tree of each node inspired hop count N is set to be 3 according to the loaded
above(each node’s SPT describes the minimum path cost data; Firstly, to calculate SPT of the nodes in and the
from itself, the root node, to all other nodes ). The second node s (total +1 nodes)using Dijkstra algorithm, and judge
step is on the basis of the first step, using heuristic search that whether there exist nodes in which are not included
theory, sets the appropriate "inspired hop count" according in the above shortest path tree, if do exist, then the path
to the computational complexity of the problem(the more meeting the requirements can’t be find; If do not exist, then

637

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.
start to calculate. Because the inspired hop count N = 3,
that is, the partial optimal path of each calculation covers 3
0 2 3
intermediate nodes in , so the first step of the calculation 0,1
is as follows: 1,2 2,1 3,3 5,1 4,1 6,1
= + +··· + 1
When = 3 :
= + + 2 3 1 3 1 2
Where represents the path weights, 0 、 1 、 2 、 Figure 4. SPT of the node 0,2,3 in the topology graph shown in figure1

3 are node numbers, = ( 1, 2) ,

1 2 Each shortest path tree above contains intermediate node 2,
representing the weight of the shortest path from the node
3, and starting node 1.
to the node .In the first cycle: = 0, ∈ , ∈
Step 3:Set inspired hop count N = 3,and λ = 3;Then
− { }， ∈ − { , }; According to permutations calculate:
and combinations available, there are ( − 1)( − 2) = + +
kinds of values of , we need to remove paths with a
Where = 0, ∈ {2,3}, ∈ {2,3} − { }, = 1; So
repeating node, then compare the weights of the remaining
there are total 2 ∗ 1 possible values for :
paths, before taking the first λ-hop nodes of the path with
(1) = + + = 2 + 1 + 1 = 4;
the minimum path weight as the next λ-hop nodes, where
λ ≤ N = 3. Might as well take λ = 1, that is, we take the (2) = + + = 1 + 1 + 3 = 5;
first hop node (node number is ) of the path with least The corresponding LinkID of the two paths are 1|5|4 and
path weight as the next hop node of the final target path in 2|6|3 respectively. Take the path with the least path
each cycle; After the first cycle, we have to make = , weight,that is, 1|5|4 is the optimal path whose traversal
repeat the above process to obtain the next hop node again; order of nodes is 0→2→3→1.
So after a total of S cycles(because there are S nodes in the
intermediate node which means S partial destination nodes), 3.3. Algorithm Analysis
we can obtain the final full path under the circumstances
that there is no ‘dead cycle’(that is, there exists no path after In the algorithm, the value of the hop count N has great
removing paths with a repeating node), or the path can’t be influence on the performance of the algorithm. The bigger
find. If there appears a ‘dead cycle’, we need to replace the the N is, the greater the possibility of appearing death cycle
current hop node(the starting node of the partial path)with in the calculation will become and the longer the time for
a suboptimal node and continue; If we have alternately computing will be, but the path weight will be smaller,
taken every optional node as the current hop node in a more closing to the optimal path’s weight. Theoretically,
certain cycle and do not find one satisfing the conditions when N = S (where S is the number of the nodes in
for the existence of a path, we need to go back to last cycle, intermediate node set), it can be guaranteed that the path
and take a suboptimal node to change the last hop node and we acquired is the optimal path, that is, the path’s weight is
calculate again. Repeat the above process until you can find the smallest.But we need to compare s! (factorial of s)
a path to reach the termination node. weight value which will take great computational effort.
Therefore, when the number of intermediate nodes(nodes
3.2. Algorithm Demonstration in intermediate node set) are not big(less than 10), we can
make N = S and λ = N. When there are many intermediate
Take the sample topology graph shown in Figure 1 for nodes, for example 20, we can no longer make N = S,
example, we are going to give a demonstration of the considering the computational effort. The specific value of
algorithm as following: the N can be determined according to the experimental
Step 1:Load data, starting node number is 0, results. In general, in order to decrease the weight of path,
termination node number is 1, intermediate node set = the value of N is generally proportional to the value of the
{2,3}, = {0,1,2,3}; In the topology of Figure 1, we can S. As for the value of λ, the greater the value of N is, the
obtain the following information of the directed graph: closer to N the λ should be, and when N = S, the λ should
0,0,1,1 be equal to N. The smaller the value of λ is, the longer the
1,0,2,2 calculation time will be, and generally speaking, the
2,0,3,1 smaller the weight of the path will be.
3,2,1,3 Assuming that the number of nodes in the node set V
4,3,1,1 is , the number of nodes in the intermediate node set V′ is
5,2,3,1 S. The time complexity of Dijkstra algorithm is ( ) (T
6,3,2,1 is the number of nodes in V), unrelated to the number of
Note: Each line is ‘LinkID,SourceID,DestinationID,Cost’ sides in directed graph. The above algorithm we proposed
Step 2: Use Dijkstra algorithm to calculate the shortest is based on Dijkstra algorithm, so the time complexity of
path tree of nodes 0,2,3, as shown in Figure 4: the algorithm is related to the number of nodes in V, and is
also related to the number of nodes in intermediate node set
V .Therefore, the time complexity of the algorithm can be

638

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.
expressed as ( ∗ S!) when N = S, which is easy to see the longest time. However, when there are more nodes in
the time complexity of the algorithm is not sensitive to , the intermediate node set(more than 20), we can no longer
the total number of nodes in V, but very sensitive to S ,the set N = S(S ≥ 20).
number of nodes in intermediate node set V .If we take
N (the insired hop count) into consideration, the time 22
!
complexity should be expressed as ( ∗ ( )!).
11 21
3.4. Program Simulation

Programming environment: Windows7 system, 2.5GHz i7- 1 0 3 4 23

4710HQ CPU, 8GB of memory, the programming
language is java and matlab, compiler environment is 9 24 5 20 26 25
Eclipse and Matlab2012a.

3.4.1 Experiment 1: Track program solution process, to 10 2 19

verify the correctness of the algorithm results
In order to facilitate tracking and observing the the 6 7 18
execution of the program, and visually verify the
correctness of the final result, this experiment selected a
sparse networks with total 27 nodes to be tested, we focused 12 8 16 14
on the effects of parameter N (inspired hop count) on the
search results of the algorithm.
13 17
Network topology data is shown in figure 5. Each line
of information is:
LinkID, SourceID, DestinationID, Cost 15
Figure 6. Weighted directed graph of the sparse network (weight is not
1,0,3,19 16,5,23,18 31,13,8,3 46,21,4,2 marked)
2,0,9,1 17,6,12,4 32,14,16,17 47,21,23,7
3,0,11,8 18,6,24,2 33,15,17,4 48,22,21,6 Table 1. Results of key steps in experiment 1
4,0,24,2 19,6,7,9 34,16,2,2 49,23,20,14
5,1,0,3 20,7,8,1 35,16,17,6 50,23,26,1 Key Step Results
6,2,5,2 21,7,2,3 36,16,7,13 51,24,3,1 Generate SPT by Nodes in and starting node
7,2,16,2 22,7,17,1 37,16,14,7 52,24,7,8
8,2,20,3 23,8,13,5 38,17,16,5 53,24,0,3 Dijkstra algorithm are all included in each SPT
9,2,19,7 24,9,0,1 39,17,15,3 54,24,6,8
10,3,24,4 25,9,10,3 40,18,19,8 55,24,5,3 Set N = 1, and Path:0→24→6→7→2→20→
11,3,4,1 26,10,6,1 41,18,26,11 56,25,26,3 26
execute algorithm
12,4,5,22 27,11,0,8 42,19,18,1 57,26,25,12
13,4,23,1 28,12,6,3 43,20,5,4 58,26,18,11 Weight=37 Time=1
14,5,2,5 29,12,13,2 44,20,26,12
Set N = 2, and Path:0→9→10→6→24→7→2
15,5,24,4 30,13,12,2 45,21,22,3
Figure 5. Topology data of a sparse network execute algorithm →20→26
Weight=33 Time=2
The network topology graph formed by the topology
data shown in Figure 5 is shown in Figure 6. Set N = 3, and Path:0→9→10→6→24→7→2
The starting node number is 0, the termination node →20→26
number is 26, and the intermediate node set = execute algorithm
{2,6,7,20,24} . The key solution steps and some Weight=33 Time=6
intermediate results are given in Table 1. We take different
Set N = 4, and Path:0→9→10→6→24→7→2
values for N (inspired hop count) and compare the final
path and the search time, where λ = 1. execute algorithm →20→26
As can be seen from the above experiment, when N = Weight=33 Time=24
1, the computing time is minimal, but the weight of path
we acquired is the largest; when N = 2, the computing Set N = 5, and Path:0→9→10→6→7→2→20
time increases exponentially, in the same time we obtained →5→24→3→4→23→26
execute algorithm
a path with a smaller weight; when N = 3,4, the calculation
time is still increasing, but come to the same path weight; Weight=32 Time=120
when N = 5 , we obtained the optimal path with the
minimum weight, but the computation time is the most.
Therefore, when N = 5, the resulting path is optimal, with

639

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.
3.4.2 Experiment 2: The effects of parameter on the path network in 3D," 2014 International Conference on Data Mining
weight and search time of the results and Intelligent Computing (ICDMIC), New Delhi, pp. 1-5, 2014.
In experiment 2, we use a more complex topological graph [2]Y. Liu, R. Song and R. Bucknall, "A practical path planning
with 20 intermediate nodes to test our algorithm. The test and navigation algorithm for an unmanned surface vehicle using
results are graphically presented in Figure 7. the fast marching algorithm," OCEANS 2015 - Genova, Genoa,
pp. 1-7, 2015.
[3]Chunbao Wang et al., "Path planning of automated guided
vehicles based on improved A-Star algorithm," 2015 IEEE
International Conference on Information and Automation,
Lijiang, pp. 2071-2076, 2015.
[4]J. Lee, B. y. Kang and D. w. Kim, "Fast genetic algorithm for
robot path planning," in Electronics Letters, vol. 49, no. 23, pp.
1449-1451, Nov. 7 2013.
[5]T.A. Bahtiyar, F. Ardilla, B.S. Marta and D. Pramadihanto,
"Effectiveness of bicycle path planning method and pure pursuit
method on omni-directional mobile robot," 2015 International
Conference on Control, Electronics, Renewable Energy and
Communications (ICCEREC), Bandung, pp. 91-97, 2015.
[6]L. Cheng, C. Liu and B. Yan, "Improved hierarchical A-star
algorithm for optimal parking path planning of the large parking
Figure 7. The effects of parameter N on the results lot," 2014 IEEE International Conference on Information and
Automation (ICIA), Hailar, pp. 695-698, 2014.
From the results of experiment 2, we can see that with the [7]J.R.Jiang, H.W.Huang, J.H.Liao and S.Y.Chen, "Extending
increase of the value of N , the search time increases Dijkstra's shortest path algorithm for software defined
exponentially, while the weight of the path is getting lower networking," 2014 16th Asia-Pacific Network Operations and
and lower. Therefore, the actual value of N depends on Management Symposium (APNOMS), Hsinchu, pp. 1-4, 2014.
your requirements on search time and path weight [8]Mingjun Wei and Yu Meng, "Research on the optimal route
respectively. choice based on improved Dijkstra," 2014 IEEE Workshop on
Advanced Research and Technology in Industry Applications
4. CONCLUSION (WARTIA), Ottawa, ON, pp. 303-306, 2014.
[9]F. H. Tseng, T. T. Liang, C. H. Lee, L. D. Chou and H. C. Chao,
In this paper, we have studied a class of complex path "A Star Search Algorithm for Civil UAV Path Planning with 3G
planning problem: In the weighted directed graph, paths Communication," 2014 Tenth International Conference on
need to be found which start from a specified node, after Intelligent Information Hiding and Multimedia Signal Processing
passing through some specified intermediate nodes without (IIH-MSP), Kitakyushu, pp. 942-945, 2014.
repetition (the number of intermediate nodes is uncertain), [10]Tan Baocheng, Wang Pei, Improvement and Implementation
and finally reach the termination node. We expect to find of A* Path Planning Algorithm[J]. Journal of Xi'an Technological
such a path with a minimum path weight within a certain University, 04:325-329, 2012.
time. To give the solution, this paper took the Dijkstra
[11]J. Yao, C. Lin, X. Xie, A. J. Wang and C. C. Hung, "Path
algorithm as the foundation, and drew on the heuristic
Planning for Virtual Human Motion Using Improved A* Star
search theory of A* algorithm, finally proposed a feasible Algorithm," 2010 Seventh International Conference on
algorithm. In this algorithm, we put forward a new concept Information Technology: New Generations (ITNG), Las Vegas,
of "inspired hop count" which is the most important NV, pp. 1154-1158, 2010.
parameter of the algorithm to dynamically adjust the
[12]Yu Haicong, Lu Feng. A Multi-modal Multi-criteria Route
balance between the path weight and the searching time.
Planning Method Based on Genetic Algorithm[J]. Journal of
Finally, the experimental results were given to show that Surveying and mapping, 01:89-96, 2014.
this algorithm can effectively solve the above path planning
problem and analyse the influence of parameter [13]Shi Enxiu, Chen Minmin, Li Jun, Huang Yumei. Research on
N (‘inspired hop count’) on the performance of the Method of Global Path-planning for Mobile Robot Based on Ant-
colony Algorithm[J]. Transactions of the Chinese Society for
algorithm.
Agricultural Machinery, 06:53-57, 2014.
5. ACKNOWLEDGMENTS [14] M. Ryan, "Constraint-based multi-robot path planning," 2010
IEEE International Conference on Robotics and Automation
This research was partially supported by the National (ICRA), Anchorage, AK, pp. 922-928, 2010.
Nature Science Foundation of China (No.61373084). [15]Huang Shuli, Hu Dasha, Jiang Yuming. Algorithm for finding
shortest path which go through specified intermediate node set[J].
6. REFERENCES Computer Engineering and Applications, 11:41-46, 2015

[1]S. A. Gautam and N. Verma, "Path planning for unmanned

aerial vehicle based on genetic algorithm & artificial neural

640

Authorized licensed use limited to: b:on Instituto Politecnico de Coimbra. Downloaded on April 04,2023 at 12:46:56 UTC from IEEE Xplore. Restrictions apply.