ISSN: 2995-3286

Research Article International Journal of Media and Networks

Research on Robot Path Planning Technology Combining Ant Algorithm and
DWA Algorithm
Hengyi Huang1* and Sanli Fu2
1
University of Sanya , No. 191 Xueyuan Road, Yingbin Avenue, *
Corresponding Author
Jiyang District, Sanya City, Hainan Province, China, 572022 Hengyi Huang, University of Sanya , No. 191 Xueyuan Road, Yingbin
Avenue, Jiyang District, Sanya City, Hainan Province, China, 572022.
2
University of Sanya , No. 191 Xueyuan Road, Yingbin Avenue,
Jiyang District, Sanya City, Hainan Province, China, 572022 Submitted: 2024, Feb 21; Accepted: 2024, Mar 11; Published: 2024, Mar 26

Citation: Huang, H., Fu, S. (2024). Research on Robot Path Planning Technology Combining Ant Algorithm and DWA
Algorithm. Int J Med Net, 2(3), 01-09.

Abstract
Path planning is an important problem in the field of robotics, the main purpose of which is to plan the best path to avoid
obstacles during the robot's action, so as to realize the robot's autonomous action. However, the traditional path planning
algorithm often cannot deal with the situation of multiple dynamic obstacles, which makes the application of the robot in the
actual scene is greatly limited. To solve this problem, a path planning algorithm based on ant colony algorithm and dynamic
window method is proposed in this paper. This fusion algorithm can not only obtain the global optimal path by using the
global known information, but also cope with the unknown dynamic obstacles to avoid collision. It can better complete the
path planning task in the actual situation with superior performance, improve the safety of the robot during the execution of
the task, and ensure the subsequent execution efficiency.

Keywords: Path planning, Ant algorithm, DWA, Fusion algorithm, Multiple Dynamic Obstacles.

1. Introduction is also the primary task of navigation in other fields, which itself
Path planning can be divided into global path planning and represents whether the intelligence of unmanned technology
local path planning [1]. The global path planning of a robot meets the requirements and determines the level of unmanned
means that the robot can search for a collision-free path from technology. As various industries begin to use different kinds of
the starting point to the target point according to static map intelligent unmanned systems, path planning algorithm, as the
information under offline conditions. In actual situations, robots core technology, needs to be developed [2].
usually acquire the surrounding environment information based
on the perception system on the robot, such as radar, sensors 2.1 Ant Algorithm Research
and other devices. In the real environment, robots often need Ant colony algorithm is a bionic algorithm that simulates the
to enter some unknown environments to perform tasks, and foraging behavior of ants in nature. It was first proposed in
may encounter static or dynamic obstacles. In order to enable 1991 by Italian scholar M.Dorigo et al. [3]. Inspired by the
robots to safely and smoothly reach the target, Some local ant's foraging behavior, he applied his ideas to the traveling
path planning algorithms are needed to enable robots to avoid salesman problem (TSP) and proposed the ant colony algorithm.
obstacles dynamically, safely and reliably when encountering This algorithm has the following advantages: strong stability,
these obstacles, and to avoid obstacles by avoiding dangers like better optimization ability and easy parallelization. At the
humans. Therefore, it is of great significance to study a general same time, it has the following disadvantages: the algorithm
path planning algorithm that conforms to the characteristics of convergence is slow, it is easy to precocious, fall into local
tasks in the real environment for robotics technology and the optimal, and the algorithm may stall in some cases. Although
strategy of strengthening artificial intelligence robots. individual ants have very limited abilities and do not have
strong visual and auditory perception systems, the ant colony as
2. Path Planning Research a whole exhibits very efficient executive ability and strong self-
Path planning refers to finding feasible motion paths that meet the organization characteristics. Through long-term observation
required standards according to the starting point and end point and analysis of ant colony behavior, relevant researchers found
of the environment in a specific environmental space, and finding that the fundamental reason why ant colonies can show certain
feasible path solutions that meet the restricted requirements. In intelligence is that ants indirectly complete communication
addition, for feasible routes, obstacles can be avoided and the between individuals through a chemical substance called
target destination can be reached safely and smoothly. As a key pheromone [4]. In the iteration process, the pheromone on the
issue in the field of unmanned systems research, path planning path will change as the ant moves. For a single ant, choosing
Int J Med Net, 2024 Volume 2 | Issue 3 | 1
the next path according to the level of pheromone concentration planning along the global optimal path direction, and finally
is generally a positive feedback mechanism. In 1989, Goss realizes the robot to effectively avoid dynamic obstacles and
and other researchers further studied the foraging behavior of plan the global optimal path.
ants through the famous "double bridge experiment" [5], thus
discovering the mystery of pheromones. 3. Ant Algorithm
Ant colony algorithm is a heuristic algorithm that simulates
2.2 DWA Algorithm Research ant foraging behavior. It finds the optimal path based on the
Dynamic window method (DWA) is relatively strong in real- pheromone trail left by ants when they search for food. Classic
time and has certain advantages in the amount of computation ant colony algorithms usually include two processes: pheromone
in moving machines Human local path planning is widely used. updating and route selection. In the process of pheromone
The DWA method was developed by Dieter Fox and Wolfram renewal, ants will leave pheromone on the path, and the amount
Burgard In 1997, et al. proposed that the general obstacle of pheromone is proportional to the quality of the path. During
avoidance algorithm did not take into account the kinematic the route selection process, ants will choose the path with higher
performance of the robot, but DWA considered the kinematic pheromone concentration for walking.
constraints of the robot, and then they successfully applied it
to the obstacle avoidance of mobile robots [6]. Compared with 3.1 Introduction to Ant Algorithm
other local path planning algorithms, DWA takes the kinematics Real ants, in search of a food source, release an ant-specific
constraints of the robot into account, and the obtained velocity secretion, often called a pheromone, along their path, which other
and other related physical information can be directly handed ants within a certain range can detect and thus influence their
to the robot's bottom control layer for control, which can be future behavior. Ants choose routes with high concentrations
well combined with path optimization and operation feedback. of pheromones, and when more ants pass through some routes,
Seder M et al. combined D* algorithm with DWA, regarded they leave higher concentrations of pheromones, which leads to
dynamic obstacles as moving grids, predicted the collision leeches.
point between obstacles and mobile robots, and then generated
virtual obstacles according to the collision point. Finally, the The probability of ants choosing this path increases, showing
two
6 algorithms
Page 6 of 18were used to achieve obstacle avoidance in positiveJournal feedback of information.
of Intelligent & RoboticUnderstood
Systems as(2024)
an enhanced
110:6
dynamic environment. Literature combined DWA algorithm learning system, ants can eventually discover the shortest path.
obstacle avoidance of mobile robots. Compared with other an enhanced learning system, ants can eventually discover
with reinforcement learning, added new evaluation function At the same time, pheromones are aromatic substances that
local path planning algorithms, DWA takes the kinematics the shortest path. At the same time, pheromones are
terms to expand
constraints of thetherobot
DWAinto evaluation
account, function
and theto obtained
address the evaporate, and their
aromatic strength diminishes
substances over time.
that evaporate, andInspired by
their strength
defect that too few factors were taken into
velocity and other related physical information can be account, and then the fact that the shortest path between Ant nests and food
diminishes over time. Inspired by the fact that the shortestsources
learned the
directly parameters
handed to theofrobot's
DWA adaptively throughlayer
bottom control Q-learning,
for can be found during theAnt
path between foraging
nestsprocess
and food of ant colonies,
sources canItalian
be found
control, which can be well combined with path
so that the agent could adapt to different location environments. scholar during the foraging process of ant colonies, Italian scholar
M. Dorigoet al. proposed a new analog evolutionary
optimization
DWA performsand operation
better than other feedback.
local pathSeder
planning M algorithms
et al. algorithmM. called
Dorigo[9] et al. proposed
Ant System a new
(AS) in 1991. analog
Here, evolutionary
Figure 1 is
combined
due to its low D* algorithm
computationwith and DWA [7], regarded
good stability, widely used to algorithm
and is dynamic illustrate thecalled Ant System
path search principle(AS) in 1991. Here,
and mechanism of antFigure 1
obstacles
used in games,as moving
mobilegrids,
robot predicted
navigation the and collision
other fields. The coloniesis[9].
point used to illustrate the path search principle and
between obstacles
biggest advantage of and
this mobile
algorithm robots,
is that and theninto
it takes generated
account mechanism of ant colonies.
virtual obstacles according to the collision point. Finally, Suppose there are two paths around the obstacle leading
the kinematic constraints of the robot to generate a smooth and Suppose there are two paths around the obstacle leading from
the two algorithms were used to achieve obstacle from the ant's Nest to the Food source: Nest-abd-food and
round safe collision
avoidance in dynamicfree path, and its output
environment. is the corresponding
Literature [8] the ant'sNest-acd-food,
Nest to the Foodwith source: Nest-abd-food
lengths of 4 and 6,andrespectively.
Nest-acd- Ants
linear speed and angular speed, which
combined DWA algorithm with reinforcement learning,can be used directly as a food, with
can travel a unit length of distance per unit ofa time,
lengths of 4 and 6, respectively. Ants can travel unit startin
control new
added instruction [8,7]. function terms to expand the DWA length of
evaluation withdistance per unit of time,
no pheromones left starting
on any with no roads.
of the pheromones
evaluation function to address the defect that too few left on any of the roads.
factors were on
2.3 Research taken intoAlgorithm
Fusion account, and then learned the
parameters
It can be seen of from
DWAtheadaptively through
above literature thatQ-learning, so that
the global static
the
pathagent
planningcould adapt to
algorithm candifferent
effectively location
plan theenvironments.
global optimal
DWA
path byperforms
using thebetter
globalthan other local
information, but path planning
it cannot avoid the
algorithms due to its
dynamic obstacles. low dynamic
Local computation path and goodalgorithm
planning stability,
and is widely used in games, mobile robot navigation and
can plan
other pathsThe
fields. online in realadvantage
biggest time and avoid of thisdynamic obstacles
algorithm is that
effectively, but because of the lack of global
it takes into account the kinematic constraints of the robot information, it is
easy
to to fall into
generate the localand
a smooth optimal
round solution
safe and often cannot
collision reachand
free path,
the target point. For these two Many scholars
its output is the corresponding linear speed and angular have combined
speed,
the twowhich
kinds ofcan be used and
algorithms directly as a good
achieved control instruction.
results in real-
time dynamic path planning. In order to plan the global optimal
2.3 Research on fusion algorithm
path in dynamic environment, a global dynamic path planning
It canbased
algorithm be seen
on antfrom
colonythe algorithm
above literature
and dynamic that the global
window
static
methodpath planninginalgorithm
is proposed this paper. can Thiseffectively
paper takes plan the
the turning
global optimal path by using the global information, but it
point of the global optimal path planned by ant colony algorithm
cannot avoid the dynamic obstacles. Local dynamic path
as the subpoint
planning of the can
algorithm dynamic
planwindow method,
paths online inand
realguides the
time and Figure 1: Schematic Diagram of the Ant Algorithm
Fig.1 Schematic diagram of the ant algorithm
dynamic window method to carry out real-time
avoid dynamic obstacles effectively, but because of the dynamic path
At t=0, the first group of 20 ants moved from the nes
lack of global information, it is easy to fall into the local
Int J Med Net, 2024 to A, and since there were no Volume
pheromones
2 | Issueon
3 | all
2 roads,
optimal solution and often cannot reach the target point. they chose either the left (ABD) or the right (ACD) road
For these two Many scholars have combined the two kinds with equal probability, so 10 ants went left (ABD) and 10
of algorithms and achieved good results in real-time went right (ACD). At the 4th unit time, the ants that
dynamic path planning. In order to plan the global optimal walked left (ABD) to the food source will double back, an
path in dynamic environment, a global dynamic path the ants that walked right (ACD) will reach the middle
At t=0, the first group of 20 ants 6movedPagefrom 18nest to A, and the path planning task of surface unmannedJournal
6 ofthe vehicleofprocess is & Robo
Intelligent
since there were no pheromones on all roads, they chose either as follows:
the left (ABD) or the right (ACD)  road Stepwith2. equal
Go through all the
probability,  ants Stepand start thethe
1. Initialize search
starting position and end point, set
process. Take ant k as an example, a tabu
so 10 ants went left (ABD) and 10 went right (ACD). At the 4th parameters such as the total number table is of iterations of the It can be seen from
ant colony
constructed according to the size of the map to
unit time, the ants that walked left (ABD) to the food source will algorithm N max, the number of iterations N
record path release
c, the numbermore of phero
the positions that ant k has already traveled, and the
double back, and the ants that walked right (ACD) will reach the ants m, and the volatilization coefficient of pheromone, accumulated pheromon
etc., and
ant is placed at the starting point to select the next
middle point of CD. At the fifth unit time, the two groups of ants raster the map. gradually increases, and
feasible grid point according to the state transition
will meet at point D. At this point, the number of pheromones  Step 2. Go through all the ants and startpath also process.
the search increases. Fin
probability formula. The path and path length of the on the best path under t
on BD is the same as on CD, because 10 ants each choose the Take ant k as an example, a tabu table is constructed according
ant are recorded during the ant's movement, until the the corresponding
corresponding path, so five returningant antshas
willnochoose
nextBD and to
feasible gridthepoint
size oforthe
hasmap to record
reached thethe positions that ant k has solut
the other five will choose CD, and the end
rightpoint,
side (ACD)
and ants
thenwill already is
the search traveled,
stopped.and Start is placed at the starting to
the antyour problem be to
point optimize
continue to move towards food. At thesearcheighth unit
withtime,
the the firstant select
next k +1.the next feasible grid point according to the state transition
five ants will return to the nest, and there will be five ants at the probability formula. The path and path length of theant
3.4 The ant algorithm
are
midpoint of AC, five ants at the Step of3.CD,
midpoint Whenand all
fiveants
ants complete the search,
recorded during that
the ant's is,
movement, until the ant has no next
afterfive
at point B. At the ninth unit time, the first the end
ants of a single
returned iteration,
to A feasible gridthe pheromone
point or has reached the end point, According
and then the to the p
number value
and were again faced with the choice of going left or right. of the path taken by the ants is algorithm
search is stopped. Start your search with the next ant k +1. has the follow
calculated according to the pheromone calculation
 Step 3. When all ants complete the search,  thatArtificial
is, after theants are
formula, and the pheromone update is carried out.
At this point, the number of tracks on AB is 20 and the number on end of a single iteration, the pheromone number In theofinitial
value the stage
AC is 15, so more ants will choose to Step
go left,4.thus
Check whether
enhancing the thepathnumber
taken by oftheiterations
ants is calculated according to theeach path is the sa
pheromone
pheromone of the route. As the process reaches
continues,thetheupper limit. calculation
difference If the number formula,of and
iterations random
the pheromone update is carried out.search in t
does not reach the upper limit,
in the number of pheromones on the two paths will get bigger  Step 4. Check whether Nc =Nc +1, andthegonumber
to of iterationsresulting
reaches the in a large
and bigger until the vast majority ofStep ants 2.
choose the shortest upper limit. If the number of iterations does not reach initial stage of the
the upper
path. This is how ants find the optimal
 Step route5. from the nest
Output the the limit,
to optimal Nc =Nc
path saved +1,during
and go tothe Step 2. convergence speed
food source when foraging. algorithm. The algorithm  isStep
complete.
5. OutputThe operation
the optimal path saved during
 the Thealgorithm.
traditional ant
flow chart of ant algorithm The is shown is
algorithm in complete.
Figure 2.The operation flow direction chart of ant search, w
3.2 Ant Algorithm Flow algorithm is shown in Figure 2. selecting the next
Path planning task, the basic ant colony algorithm in solving artificial ants to so
path length obtain
optimal path, and
not conducive to t
unmanned boats, a
robot motor and o
completely follow
of the motor and o
time, in the actual
can move at any A
any heading Angle
 In practical applic
to defects such as
solution and deadl
efficiency and con
deteriorate, and th
local extreme valu
gradually increase

4 DWA algorithm
The dynamic window
Figure 2: Diagram of the Ant Algorithm
algorithm, which abstra
into a window and find
3.3 Ant Algorithm Experiment Simulation Fig.2 diagram of the ant algorithm
Common local obstacle
The planned path of the ant algorithm simulation experiment is shown in Figure 3.
3.3 Ant algorithm experiment simulation robots include artificial
The planned path of the ant algorithm simulation window method (DWA
experiment is shown in Figure 3. control and so on. Amo
Int J Med Net, 2024 is relatively
Volume 2 | Issue 3 | strong
3 in re
in computational amoun
planning of mobile robo
 C
3.3 Ant algorithm experiment simulation ro
The planned path of the ant algorithm simulation w
experiment is shown in Figure 3. co
is
in
pl

4.

sa
an
Figure 3: Ant Algorithm Simulation Diagram sa
Fig.3 Ant algorithm simulation
It can be seen from Figure 3 that ants with a shorter path release
the algorithm diagram
gradually deteriorate, and the probability of ants ev
more pheromone. As time goes on, the accumulated pheromone falling into local extreme value points and deadlock points
concentration on the shorter path gradually increases, and the gradually increases.

②…
number of ants choosing this path also increases. Finally, the
whole ant will concentrate on the best path under the action of
positive feedback, and the corresponding solution is the optimal
4. DWA Algorithm
The dynamic window method is a local path planning algorithm,
solution to the problem to be optimized. which abstracts the obstacles around the robot into a window
and finds a feasible path within the window. Common local
3.4 The Ant Algorithm has Problems obstacle avoidance algorithms for mobile robots include artificial
According to the previous simulation, the ant algorithm has the potential field method, dynamic window method (DWA),
following shortcomings: velocity obstacle method, fuzzy control and so on. Among them,
 Artificial ants are blind in the initial stage of search. In dynamic window method is relatively strong in real-time and
the initial stage, the pheromone concentration on each path is has certain advantages in computational amount, and is widely
the same by default, so ants conduct blind random search in used in local path planning of mobile robots [10].
the initial stage of the algorithm, resulting in a large number of
random paths in the initial stage of the algorithm, which affects 4.1 DWA Algorithm Introduction
the convergence speed of the algorithm. The idea of DWA algorithm is to collect robot velocity samples
through the robot mathematical model, predict and simulate
 The traditional ant colony algorithm adopts eight-direction the motion trajectories generated under the sample velocity for
search, which will limit the Angle of selecting the next path a period of time, conduct standard evaluation on these motion
node in the search process of artificial ants to some extent, and trajectories, select a group of optimal trajectories, and the robot
finally make the path length obtained by the algorithm exceed moves according to the optimal trajectories. The motion attitude
the ideal optimal path, and the path Angle is too large, which is and direction of the robot are determined by the current linear
not conducive to the actual tracking and movement of unmanned velocity and angular velocity (turning velocity) of the robot.
boats, and has higher requirements for the robot motor and other
actuator. If the robot completely follows this path, it may cause DWA algorithm mainly includes three steps: velocity sampling,
the failure of the motor and other power equipment; At the same trajectory prediction (calculation), trajectory evaluation. In the
time, in the actual navigation environment, the robot can move velocity vector space Vr, the continuous velocity vector space VR
at any Angle, that is, the path is a path with any heading Angle. is discretized according to the sampling points of linear velocity
and angular velocity, and the discrete sampling points (v,ω) are
 In practical application, ant colony algorithm is prone to obtained. For each sampling point, multiple motion trajectories
defects such as low search efficiency, local optimal solution and of the robot in the next moment are predicted according to the
deadlock. Therefore, the search efficiency and convergence of robot kinematics model, as shown in Figure 4.

Int J Med Net, 2024 Volume 2 | Issue 3 | 4

 and the discrete sampling points (v,ω) are obtained. For
each sampling point, multiple motion trajectories of the
robot in the next moment are predicted according to the
robot kinematics model, as shown in Figure 4.

Fig5 Flowchart of DW
Figure 4: DWA Algorithm Schematic Diagram
Fig4 DWA algorithm schematic diagram
4.2 DWA Algorithm Flow set simulation cycle time. 4.3 DWA algorithm emula
windowalgorithm
4.2 DWA
The algorithm steps of dynamic flow
method are as follows:
 Step 5. After considering the position of the target point,
 Step 1. Initialize the start position and end position. select the current optimal trajectory and the corresponding speed
The algorithm steps of dynamicpair window
based onmethod are as function.
the evaluation
 Step 2. Calculate the feasible
follows:velocity space, the dynamic
window.  Step 6. The robot uses the optimal speed pair in Step 5 to
move.
 Step 3. Sample the dynamic Step 1.obtained
 window Initialize the start
in Step 2 and position and end position.
obtain the set of speed pairs according to the set speed resolution.  Step 7. Determine whether the robot has reached the target
 Step 2. Calculate the feasiblepoint velocity
at thisspace,
time. Ifthe
it has reached the target, output the global
 Step 4. For the speed pair setdynamic
obtained window.
Journal of Intelligent
above, traverse&the
Robotic Systemsand end
trajectory (2024) 110:6
the algorithm; Otherwise, go to Step 2. The
robot trajectory corresponding to the speed pair according to the algorithm flow chart is shown in Figure 5 below:
d the robot moves according to the  Step 3. Sample the dynamic window obtained in Step
e motion attitude and direction of 2 and obtain the set of speed pairs according to the set
d by the current linear velocity and speed resolution.
ng velocity) of the robot.
 Step 4. For the speed pair set obtained above, traverse
y includes three steps: velocity the robot trajectory corresponding to the speed pair
ediction (calculation), trajectory according to the set simulation cycle time.
city vector space Vr, the continuous
 Step 5. After considering the position of the target Fig6 Dwa Algorithm
VR is discretized according to the point, select the current optimal trajectory and the
ar velocity and angular velocity, corresponding speed pair based on the evaluation It can be seen from FIG. 6 th
function.
ng points (v,ω) are obtained. For information of dynamic obstacl
ultiple motion trajectories of the Step 6. The robot uses the optimal speed pair in Step robot can predict the trajectory
nt are predicted according to the 5 to move. whether they collide with the ro
l, as shown in Figure 4. that may collide with the robot,
 Step 7. Determine whether the robot has reached the
target point at this time. If it has reached the target, dynamic obstacles on the simul
output the global trajectory and end the algorithm; evaluated through the evaluatio
Otherwise, go to Step 2. The algorithm flow chart is
shown in Figure 5 below: cost function. This allows the r
in time to avoid dynamic obstac

②…
Figure 5: Flowchart of DWA Algorithm

4.3. DWA Algorithm Emulation

Fig5 Flowchart of DWA algorithm
lgorithm schematic diagram
4.3 DWA algorithm emulation
flow
Int J Med Net, 2024 Volume 2 | Issue 3 | 5

of dynamic window method are as

 Fig5 Flowchart of DWA algorithm
tic diagram
4.3 DWA algorithm emulation

dow method are as

and end position.

city space, the

ow obtained in Step
according to the set

ined above, traverse

to the speed pair
cle time. Figure 6: Dwa Algorithm Simulation Diagram

tion of the target

It can be seen from FIG. 6 that after Fig6sensing
Dwa theAlgorithm
motion thesimulation
robot's actionsdiagram
by adjusting the pheromone concentration
ajectory and the
information of dynamic obstacles through sensors, the robot can and window size.
predict the trajectory of obstacles and determine whether they
n the evaluation It can obstacles
collide with the robot. For dynamic be seenthat from mayFIG.
collide6 that after sensing
5.1 Advantages the Algorithm
of Fusion motion
information
with the robot, the repulsive force of dynamicof dynamic
obstacles on obstacles
the In the through
current sensors, the tasks related to robot path
research project
simulated trajectory can be evaluated through the evaluation planning, firstly, the optimal path is obtained offline according
speed function
pair in inStep
the DWA cost robot canThis
function. predict
allowsthethetrajectory
robot to toof theobstacles and determine
global path planning algorithm based on the known static
turn or slowdown in time to avoid dynamic obstacles in the environment
whether they collide with the robot. For dynamic obstacles information, that is, the global path planning has
environment. been completed before the mobile robot executes the actual task,
that may collide with the robot,butthe the repulsive
disadvantageforce
is that of
such planning cannot cope with the
bot has 4.4
reached the
Here are Problems with DWA Algorithm
eached The target, dynamic dynamic
the traditional obstacles on the simulated trajectory can be during the robot operation. The
dynamic obstacles encountered
window method does not consider purpose of local path planning is to enable the mobile robot to
nd the algorithm; evaluated
the size of the mobile robot itself, andthrough theaffect
its size will evaluation
the copefunction
with dynamicin the DWA
obstacles in the process of moving so as to
rithm flow chart
planning is of the dynamic window method in practical safely avoid obstacles. The disadvantage of such planning is that
effect
cost function. This allows the robot to turn or slow down
application. At the same time, the traditional dynamic window it does not take into account the global map, but only considers
method only considers thein time
local to avoid
optimal dynamic
trajectory when guidingobstacles
partial in the environment.
incomplete environmental information within a certain
mobile robots such as unmanned boats to move, but there may range, which may fall into local optimization, resulting in greater
be many areas in the map that lead to robot deadlock, such as blindness and randomness of the generated path. They tend to be
U-shaped areas, which may lead to local minimum problem, poor in length or other aspects.
because the overall shape of the map is not taken into account.
The robot moves only according to the optimal trajectory of the 5.2 Fusion Algorithm Flow
current moment; In addition, in the traditional dynamic window  Initialize pheromones: Abstract the environment around the
method, when the distance between the robot and the target point robot into a grid, and initialize each grid to a pheromone value.
is small, the Angle evaluation term in the evaluation function  Ant colony algorithm process: Ants will move from the current
may change sharply, and the original evaluation function will grid to another surrounding grid with a certain probability, and
prompt the robot to increase its angular velocity to narrow the leave pheromones in the process of moving. The concentration
Angle gap. Its disadvantage is that the robot will shock when of pheromones depends on the quality of the path. The path the
approaching the target, and the robot may not reach the target ants choose is influenced by the pheromone concentration and
point at all when the situation is not ideal. the heuristic function.
 Dynamic window method process: According to the speed and
5 . Multi-Robot Fusion Algorithm direction of the robot, determine the size of the window around
In order to solve the path planning problem of multiple dynamic the robot. The grid in the window is regarded as a feasible path
obstacles, a path planning algorithm based on ant colony and the best path is selected.
algorithm and dynamic window method is proposed in this  Update the pheromone: Update the pheromone concentration
paper. The algorithm regards ants as robots and pheromones as according to the path taken by the robot, so that more pheromones
state information about the robot's surroundings, and controls accumulate in areas with higher path quality.

Int J Med Net, 2024 Volume 2 | Issue 3 | 6

 Update the pheromone: Update the pheromone
concentration according to the path taken by the
robot, so that more pheromones accumulate in areas
with higher path quality.
Journal of Intelligent & Robotic Systems (2024) 110:6 Page 7 of 18 6

5.3 Fusion algorithm simulation

5.3.1 Fusion experiment without obstructions1

In order to verify the effectiveness of the proposed path

planning algorithm, an experiment is carried out. The initial
target path planning test of the robot was carried out in the
environment without
Journal of Intelligent additional
& Robotic obstacles,
Systems and
(2024) the test results
110:6 Page 7 of 18 6
are shown in Figure 8.
5.3 Fusion algorithm simulation
5.3.1 Fusion experiment without obstructions1

In orderofto
Journal verify &
Intelligent theRobotic
effectiveness
Systems of the proposed
(2024) 110:6 path Page 7 of 18 6
planning algorithm, an experiment is carried out. The initial
target path planning test of the robot was carried out in the
5.3 Fusion
environment withoutalgorithm
additionalsimulation
obstacles, and the test results Figure 10: Local Planning Path Obstacle Avoidance Figure 1
are shown
Journal5.3.1 in Figure 8.
Fusion&experiment
of Intelligent Robotic Systemswithout obstructions1
(2024) 110:6 Page 7 of 18 6
Journal of Intelligent & Robotic Systems (2024) 110:6 Page 7 of
1 18
As can Fig10
be seenLocal
fromplanning path FIG.
FIG. 8 and obstacle avoidance
9, when Figure
the robot's
5.3 Fusion In order to verifysimulation
algorithm the effectiveness of the proposed path environment is known and there are no additional dynamic
5.3planning
Fusion
Figure 7: algorithm
algorithm,
Fusion simulation
an experiment
Algorithm
is carried out. The initial
Flow Chartout in the unknown obstacles, the robot runs global path planning, and the
5.3.1 Fusion experiment without obstructions1
target path planning test of the robot was carried
environment without additional obstacles, and the test results path convergence data is shown in FIG. 9.
5.3.1
Fig7 Fusion
Fusion
are Algorithm
shown experiment
algorithm
in Figure 8. without obstructions1
flow chart
5.3 Fusion Simulation
In order
5.3.1 Fusion to verify the
Experiment effectiveness
Without of the proposed path 5.3.2 Fusion Experiment Dynamic Obstacle 2
Obstructions1
planning algorithm,
In order an
to testexperiment
verify therobot is carried
effectiveness out.
pathThe
theininitial
was carriedofout proposed
In order to verify the effectiveness of the proposed planning Thispath
experiment is an extended version of experiment 1, the
target path planning of the the Fig10 Local planning path obstacle avoidance Figure 1
planning
algorithm,
Fig8 an
Plan algorithm,
experiment
the path is an
diagram experiment
carried out.
globallyThe is carried
initial
environment without additional obstacles, and the test results target out.
path The initial
main purpose of which is to verify the collision avoidance
aretarget
planning testpath
shown of planning
the
in Figure robot
8. was testcarried
of theout robot
in thewas carried out effect
environment in theof fusion algorithm on multi-target dynamic obstacles.
environment
without without and
additional obstacles, additional
the test obstacles, and the
results are shown in test resultsobstacles are added to the aforementioned robot path
Dynamic
Figureare
8. shown in Figure 8. environment. The simulation test of robot path planning under
multi-target
Fig10 dynamic obstacles
Local planning environment
path obstacle is shown
avoidance Figure in
1 Figure
10.
Fig8 Plan the path diagram globally

Fig10 Local planning path obstacle avoidance Figure 1

Fig8 Plan the path diagram globally

Fig10 Local planning path obstacle avoid

Fig11 Local planning path obstacle avoidance Figure 2

Fig9 Global
Fig8 Plan planning
the path path convergence diagram
Figure 8: Plan the diagram globally
Path Diagram Globally
Figure 11:Local
Fig11 Localplanning
Planning Path
path Obstacle
obstacle Avoidance
avoidance Figure Figure
2 2
As can be seen from FIG. 8 and FIG. 9, when the robot's
environment is known and there are no additional dynamic
unknown Fig8
Fig9 Plan
Global
obstacles, the the path
planning
robotpath diagram
runsconvergenceglobally
diagram
global path planning, and
the path convergence data is shown in FIG. 9.
Fig11 Local planning path obstacle avoidance Figure 2
5.3.2 As can be
Fusion seen from FIG.
experiment 8 and FIG.
dynamic 9, when the
obstacle 2 robot's
environment is known and there are no additional dynamic
Fig9 Global planning path convergence diagram
This experiment
unknown is an
obstacles, the extended
robot runs version of experiment
global path planning, and1,
the path convergence data is shown in FIG. 9.
the main purpose of which is to verify the collision avoidance
effect As algorithm
can
of environment
5.3.2 fusion
Fusion be seen on
experiment from FIG. 8 and
multi-target
dynamic FIG. 9, when
dynamic
obstacle the robot's
2 obstacles.
is known and there are no additional dynamic
Dynamic obstacles
Thisunknown
experiment are added
is an
obstacles, to theversion
extended
the robot aforementioned
runs of experiment
global robot
1, and
path planning,
the purpose
path convergence data
is toisverify
shownthein collision
FIG. 9. avoidance
paththe main
environment. of which
The simulation test of robot path planning
effect
under of fusion
multi-target
5.3.2 algorithm
dynamic
Fusion on multi-target
obstacles
experiment dynamic
environment
dynamic obstacle obstacles.
is
2 shown
DynamicThis
in Figure 10. obstacles are added to the aforementioned
experiment is an extended version of experiment robot 1, Fig11
Fig12 Local
Local planning
planning path
route endobstacle avoidance Figure 2
point diagram
Figure 9: Global Planning Path Convergence Diagram
path environment. The of
simulation test of robot path planning Figure 12: Local Planning Route End Point Diagram
the main purpose which is to verify the collision avoidance
under multi-target
effect
Int J MedFig9 of
Global
Net, 2024
dynamic
fusion planningobstacles
algorithm on environment
multi-target
path convergencedynamicis shown
obstacles.
diagram Volume 2 | Issue 3 | 7
in Figure 10. obstacles are added to the aforementioned robot Fig12 Local planning route end point diagram
Dynamic
path environment. The simulation test of robot path planning
As canmulti-target
under be seen from FIG.obstacles
dynamic 8 and FIG. 9, when
environment is the robot's
shown
Fig11 Local
Fig12 Local planning route end pointplanning
diagram path obstacle avoida
environment
in Figureis10.
known and there are no additional dynamic
unknown obstacles, the robot runs global path planning, and ②…
 6 6Page
Page 1818
18 of of 18 Journal
Journal of Intelligent
of Intelligent & Robotic
& Robotic Systems
Systems (2024)
(2024) 110:6
110:6
Fig15
Fig15 Local
Local planning
planning path
path obstacle
obstacle avoidance
avoidance mapmap

Fig13
Fig13 Fusion
Fusion planning
planning pathpath trajectory
trajectory diagram
diagram
Fig16
Fig16 Local
Local planning
planning terminal
terminal route
route planning
planning diagram
diagram
5.3.3
5.3.3 Fusion
Fusion experiment
experiment mixed
mixed environment
environment obstacle
obstacle
Fig13 Fusion planning path trajectory diagram
3 3 13: Fusion Planning Path Trajectory Diagram
Figure Fig16 Local
Figure planning
16: Local terminal
Planning route Route
Terminal planning diagramDiagram
Planning
5.3.3 FusionFig13 experiment
Fig13 Fusion
Fusion planningmixed
planning path environment
path trajectory
trajectory obstacle
diagram
diagram
ThisThis experiment
experiment anis extended
isMixed an extended version
version of experiment
of experiment Fig16
2, Fig16 Local planning terminal route planning diagram
3 5.3.3 Fusion Experiment Environment Obstacle 3 2, Local planning terminal route planning diagram
5.3.3 Fusion
the main experiment
purpose ofmixed mixed
which environment
isverify
to verify the obstacle
collision avoidance
Thisthe
5.3.3 main
Fusion
experiment purpose is anofextended
experiment which is version
toenvironment
oftheexperiment
collision
obstacle avoidance
2, the
main effect
3 experiment
effect
purpose of
of fusion
of fusion
which algorithm
algorithm
is toon on multi-target
multi-target
verify dynamic
dynamic
theexperiment
collision obstacles
2,obstacles
avoidance
3This is an extended version of
effectandofand unknown
unknown
fusion algorithm static
static on obstacles.
obstacles.
multi-target
the main purpose of which is to verify the collision avoidance Dynamic
Dynamic and
dynamic and
staticstatic unknown
unknown
obstacles and
unknown This experiment
obstacles
static
obstacles
This are
obstacles.
are added
experiment is is
added an extended
to
Dynamic
toextendedthe and version
aforementioned
static
the aforementioned ofrobot path 2, 2,
experiment
unknown robot path
obstacles
effect of fusion algorithm onanmulti-target version
dynamic ofobstacles
experiment
the
are main
added purpose
to
environment.
environment. the
ofTheof which
The is
aforementioned to
issimulation
simulation verify
test test
of the
robotrobotcollision
path
of robot avoidance
environment.
path planning
pathavoidance
planning
and the main
unknown purpose
static obstacles.which to verify
Dynamic andthe collision
static unknown
The simulation
effect of
under
effectare fusion
under
of fusion test of
algorithm
multi-target
multi-target
algorithm robot on path
dynamic
dynamic and
on multi-target planning
multi-target
and static
static under
dynamic
obstacles
obstacles
dynamic multi-target
obstacles
environment
environment
obstacles added to the aforementioned robot path obstacles
dynamic
and isand
unknown
shownstatic
static
in obstacles
obstacles.
Figure environment
14. Dynamic is shown
and static inunknown
Figure 14.
environment. is shown
and unknown in Figure
The static
simulation 14.
obstacles.
test ofDynamic
robot path andplanning
static unknown
obstacles
obstacles are are added
added to to the
the aforementioned
aforementioned robotrobot path
path
under multi-target dynamic and static obstacles environment
environment.
environment. The simulation test of robot path planning
is shown in FigureThe 14. simulation test of robot path planning
under multi-target
under multi-target dynamic dynamic andand static
static obstacles
obstacles environment
environment
is shown in Figure
is shown in Figure 14. 14. Fig17
Fig17 Fusion
Fusion algorithm
algorithm planning
planning pathpath trajectory
trajectory diagram
diagram
Figure 17: Fusion Algorithm Planning Path Trajectory
Fig17Diagram As can
Fusion algorithm As can be seen
beplanning
seen from from
pathFIG. FIG.
14 and
trajectory 14 and 15, 15,
when
diagram when therethere
are are
unknown
unknown dynamic
dynamic andand static static obstacles
obstacles in thein the environment
environment of of
AsFig17
canthebethe seen
Fusion
robot, from
robot,
when whenFIG.
algorithm 14
the
theplanning
robot and
planning
robot 15, when
path
approaches
approaches thethere
trajectorythe are
dynamic unknown
diagram
dynamic obstacles,
obstacles,
Fig17
As canFusion beandseen algorithm
fromobstacles
FIG. 14 and path trajectory
15,environment
when there diagram
are
dynamic the static
robot takes timely inobstacle
the avoidance of the robot,
measures and
unknown the robot takes timely obstacle avoidance measures of and
when dynamic
theAs robot
can
and
be
static obstacles
approaches
seen from the dynamic
FIG.
in the environment
14 and 15,obstacles,
when therethe robot
are
makes makes certain
certain adjustments
adjustments to to
its its
path. path.When When it arecertain the the
it encounters
encounters
takesAs
the robot, when can the
timely beobstacle
seen
robotfrom FIG. 14the
approaches
avoidance anddynamic
measures15, when and there
obstacles,
makes
unknown
unknown tunneltunnel dynamic
formed
dynamic formed
and and
by Whenby static
the
the avoidance
static unknown obstacles
unknown
obstacles static
in thein
static the
obstacles,environment
obstacles,
environment itof
it also of
also
the robot takes
adjustments timely
to obstacle
its path. it encounters measures the and formed
tunnel by
the
thecertainrobot,
successfully
robot,
unknown when when
successfully
static the
robotrobot
completes
thecompletes
obstacles, approaches
the
approaches the
task
it also task
of it
the the
safe dynamic
ofencounters
safe
dynamic
successfully collision obstacles,
collision avoidance.
avoidance.
obstacles,
completes the
makes adjustments to its path. When the
Figure 14: Mixed Obstacle Environment Plan tasktheof robotThis
safe
Thistakes takes
collision
experiment timely
experiment obstacle
successfully
avoidance.
successfully avoidance
This demonstrates
experiment
demonstrates measures theand performance
successfully
theformed
tunnel robot by thetimely
unknown obstacle static avoidance
obstacles, it alsotheand
measures performance
Fig14
Fig14 Mixed
Mixed obstacle
obstacle environment
environment planplan makes
demonstrates
of certain
of
the thethe
fusion adjustments
fusion algorithm.
performance
algorithm. toofitsthe path.
fusion When it encounters
algorithm. the
successfully completes the task of safe collision avoidance. the
makes certain adjustments to its path. When it encounters
tunnel formed by the unknown static obstacles, it also
Thistunnel
experimentformed by the unknown
successfully demonstratesstatic obstacles,
the performance it also
5.4successfully
Fusion Algorithm completes Innovation
the task Points
of safe collision avoidance.
Fig14 Mixed obstacle environment plan successfully
of the 5.4 Fusion
completes algorithm ofinnovation
the taskinnovation safe collision points
avoidance.
Infusion
this5.4
This
algorithm.
Fusion
paper,
experiment
algorithm
the global path planning
successfully
points
algorithm
demonstrates the and local path
performance
This experiment In In successfully
this this
paper,paper, the thedemonstrates
globalglobalpath path the performance
planning
planning algorithm
algorithm andand
Fig14 Mixed obstacle environment plan planning
of the algorithm
fusion are
algorithm. integrated, and simulation experiments
Fig14 Mixed obstacle environment plan of
arethe fusion
carried
local local
out algorithm.
path
in planning
different
path planning algorithm
environments.
algorithm are integrated,
Simulation and
experiments
are integrated, and simulation simulation
5.4 Fusion algorithm innovation points
verify the experiments
feasibility
experiments are ofarethe
carried
In this paper, the global path planning carried
algorithm.
out out
in inThe
different
different algorithm
algorithm environments.
and can make
environments.
good5.4 Fusion
use of the
Simulation algorithm
global optimal
experimentsinnovation
pathverifyand points
use
the the improved DW
local5.4
pathFusion
Simulation
planning algorithm experiments
algorithm innovation verify
are integrated, points
the andfeasibility
feasibility
simulationof the of the algorithm.
algorithm.
algorithm In to
this
The avoid
paper,
algorithm collision
the global
can with
make pathunknown
good planning
use dynamic
of algorithm
the obstacles.
global and
optimal
pathpath
experiments InThethis
arealgorithm
paper,
carriedthe can
out make
global
in good
path
different use of the
planning
environments. global optimal
algorithm and
Onlylocalbypathcombining
and planning
use the above
algorithm twoare path planning
integrated, and can the robot
simulation
local and
path use
planningthe the improved
improved
algorithm DW DW algorithm
algorithm tothe toalgorithm.
avoidavoid collision
collision withwith
Simulation
safely experiments
and
experiments efficiently
are
verify
reach the
carried theare
out
integrated,
feasibility
target
in different on and
pointofenvironments. simulation
demand.
The experiments
algorithm can unknown
unknown aremake dynamic
dynamic
carriedgood outuse obstacles.
obstacles.
in ofdifferentOnly
the global Only by by combining
combining
environments.
optimal path the the above
above
Simulation two experiments
path planning verify
can thethe feasibility
robot safely of
and the algorithm.
efficiently reach
use thetwo
and Simulation path
improved
6. Conclusions planning
experimentsDW algorithm can thethe
verify robot safely
tofeasibility
avoid and
of the
collision efficiently
algorithm.
with reach
The The algorithm
the the
Pathalgorithm target target can
point
can make
point
make on good
is demand.
on demand.
good anuse use of the
of the global global optimal path
unknown planning
dynamic algorithm
obstacles. Only byimportant
combining coretheoptimal
above path of
technology
Figure 15: Local Planning Path Obstacle Avoidance Map two and and
robot use
motion the improved
research, andDW algorithm
reasonable to
path avoid
planningcollision with
algorithm
pathuse the improved
planning can the DW robotalgorithm
safely andtoefficiently
avoid collision reach with
is unknown
of
unknown great dynamic
significance obstacles. Only by
obstacles. Only by combining the above of
to improve the combining
economy and the above
safety
the target pointdynamic
on demand.
two path planning cancan
two
robot path
operation. planning This thethe
paper robot
proposes
robot safely
safely a and
path and efficiently
planning
efficiently reach
algorithm
reach
based on ant colony algorithm and dynamic window method. In
thethe target
target pointpoint on on demand.
demand.
this paper, the proposed fusion algorithm is simulated for path
Int J Med Net, 2024 Volume 2 | Issue 3 | 8
planning under different environments with multiple dynamic affiliations.
obstacles. The test results show that the fusion algorithm can Springer Nature or its licensor (e.g. a society or other partner)
deal with path planning of multiple dynamic obstacles, and the holds exclusive rights to this article under a publishing
proposed algorithm has a good path planning effect. It has higher agreement with the author(s) or other rightsholder(s); author
robustness and better adaptability. In the future, we can further self-archiving of the accepted manuscript version of this article
explore the application of the algorithm proposed in this paper is solely governed by the terms of such publishing agreement
in the field of robots to promote the autonomous action and and applicable law.
intelligent development of robots.
Hengyi Huang, lecturer, teacher at the School of New Energy
Acknowledgements and Intelligent Connected Vehicles, Sanya University. Received
This research is supported by Sanya University Virtual Faculty a master's degree from Hubei University in 2014. His research
Pilot Construction Project (SYJZXN202303) scope includes machine learning and path planning.

Author Contributions Sanli Fu, lecturer, teacher at the School of New Energy and
The authors have worked together to complete this research. Intelligent Connected Vehicles, Sanya University. Received a
All authors contributed to the study’s conception and design. master's degree from Hubei University in 2014. Her research
Sanli Fu and Hengyi Huang performed research, analyzed the scope includes machine learning and path planning.
data and were involved in writing the manuscript.Hengyi Huang
and Sanli Fu collected, analysed, and interpreted the data from References
differ- entresearch articles to formulate a summary and to create 1. Liu, Y., & Bucknall, R. (2015). Path planning algorithm
a roadmap for future researchers. for unmanned surface vehicle formations in a practical
maritime environment. Ocean engineering, 97, 126-144.
Data Availability 2. Huang Yan, Li Yan, Yu Jiancheng, et al. Current Situation
The data presented in this study are available on request from and Development Trend of AUV Intelligence [J]. Robot,
the corresponding author. The datasets used and/or anal- ysed 2019,42(2):215-231. (in Chinese)
during the current study are available from the corresponding 3. Colorni, A., Dorigo, M., & Maniezzo, V. (1991, December).
author on reasonable request. Distributed optimization by ant colonies. In Proceedings of
the first European conference on artificial life (Vol. 142, pp.
Declarations 134-142).
The author(s) declared no potential conicts of interest with 4. Mohan, B. C., & Baskaran, R. (2012). A survey: Ant Colony
respect to the research, authorship, and/or publication of this Optimization based recent research and implementation
article. on several engineering domain. Expert Systems with
Applications, 39(4), 4618-4627.
Ethical Approval 5. Deneubourg, J. L., Aron, S., Goss, S., & Pasteels, J. M.
This article does not contain any studies with human participants (1990). The self-organizing exploratory pattern of the
or animals performed by any of the authors. argentine ant. Journal of insect behavior, 3, 159-168.
6. Fox, D., Burgard, W., & Thrun, S. (1997). The dynamic
Consent to Participate window approach to collision avoidance. IEEE Robotics &
Not applicable. Automation Magazine, 4(1), 23-33.
7. Seder, M., Baotić, M., & Petrović, I. (2016). Receding
Consent for Publication horizon control for convergent navigation of a differential
We, the authors give consent for the publication of identifiable drive mobile robot. IEEE transactions on control systems
details, which can include photographs or details within the text technology, 25(2), 653-660.
(“Material”) to be published in the above Journal Therefore, 8. Chang, L., Shan, L., Jiang, C., & Dai, Y. (2021).
anyone can read the material published in the Journal. Reinforcement based mobile robot path planning with
improved dynamic window approach in unknown
Conflicts of Interest environment. Autonomous robots, 45, 51-76.
The authors declare no conflict of interest. 9. Dorigo, M. (1991). Positive feedback as a search strategy.
Institutional Review Board Statement The study in the paper Technical report, 91-16.
did not involve humans or animals. scope includes intelligent 10. Wang Yongxiong, Tian Yongyong, Li Xuan, Li Lianghua.
robots, machine learning, and path planning. Adaptive Dynamic Window Method for Crossing Dense
Obstacles [J]. Control and Decision,2019,34(05):927-936.
Publisher’s Note Springer Nature remains neutral with regard
to juris- dictional claims in published maps and institutional
Copyright: ©2024 Hengyi Huang, et al. This is an open-access article
distributed under the terms of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are credited.
Int J Med Net, 2024 https://opastpublishers.com Volume 2 | Issue 3 | 9