Proceedings of the 2023 International Conference on Mechatronics and Smart Systems

DOI: 10.54254/2755-2721/10/20230170

Path planning algorithm based on Improved Artificial

Potential Field

Eryi Zhang
School of University College Dublin, Dublin, Ireland

[email protected]

Abstract. The domain of research and development concerning mobile robot obstacle avoidance
continues to remain an active area of interest. Artificial potential fields (APF) are a common and
effective method for obstacle avoidance path planning, where the robot is guided to the target
location by a simulated environmental potential field. Traditional artificial potential field
methods tend to trap robots in local minima, impeding their ability to reach the goal. This
research endeavours to introduce a new approach, the Improved Artificial Potential Field (IAPF)
algorithm, which incorporates the A-star method in constructing the artificial potential field. This
technique more effectively addresses the issue of path planning for mobile robots, thereby
avoiding local minimum solutions. Through simulation experiments in different scenarios, the
feasibility of the IAPF algorithm of this paper is verified. The results show that, compared with
the traditional APF method, the IAPF algorithm can solve problem of local minimum and plan
a sensible path.

Keywords: path planning, A-star, obstacle avoidance, artificial potential field.

1. Introduction
The advent of robot automation has resulted in the widespread application of mobile robots in various
fields such as manufacturing, logistics, distribution, and medical care. However, the robot needs to avoid
obstacles while automatically planning its path during its tasks, which requires specific method. The
APF method is a commonly used robot obstacle avoidance method. It was first proposed by Khatib in
1986 and applied to the collision avoidance problem of the manipulator. The fundamental principle
underlying the APF method involves constructing a potential field that reflects the environment in which
a robot operates. This potential field affects the mobile robot, allowing it to be pulled towards its target
position by a gravitational field while being repelled by the obstacle's repulsion field. The APF method
enables a mobile robot to move towards the target point by constructing a potential field based on the
environment. However, the traditional method has a tendency to get trapped in local minimum solutions,
which restricts the robot's ability to reach the desired target and hinders its practical applications.
For the local minimum problem, Scholars have proposed many improved methods for the minimum
point problem, such as randomized escape [1], heuristic search [2], harmonic potential field [3],
simulated annealing [4], semi-landmarking [5], fuzzy artificial potential field [6], extremal Optimization
[7], wall following [8], virtual flow [9] and so on. Their center ideas are treating the moving trajectory
of the robot as a state sequence or replacing the binary logic operation in the traditional method with
other operations. These methods are able to solve the local minimum question to a certain extent, but
© 2023 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(https://creativecommons.org/licenses/by/4.0/).

167
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

most of these methods are complex and computationally intensive, which is not conducive to real-time
update.
Therefore, this study aims to propose an IAPF algorithm for mobile robot path planning. The IAPF
algorithm introduces the A-star technique when constructing the potential field, which can better avoid
the emergence of local optimal solutions.

2. Artificial potential field

If a particle in any position in a space area, by the size and direction are determined by the single value
of the force, then the region is called the force field[10], the force field on the particle force F writing
space position of single value differentiable function[11] :
𝐹𝐹𝑥𝑥 = 𝐹𝐹𝑥𝑥 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧)
�𝐹𝐹𝑦𝑦 = 𝐹𝐹𝑦𝑦 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧) (1)
𝐹𝐹𝑧𝑧 = 𝐹𝐹𝑧𝑧 (𝑥𝑥, 𝑦𝑦, 𝑧𝑧)
If there is a single-valued function U = (x, y, z), its gradient is exactly equal to the force F in (1), the
following equation holds [11]:
∂U ∂U ∂U
Fx = , Fy = , Fz = or F = gradU (2)
∂x ∂y ∂z
Then this special force field in (2) is called the force field, also known as the conservative force field,
and U is called the potential function.
The APF method entails establishing an APF for the robot's operational environment. This field is
created by superimposing the gravitational field of the target location with the repulsive field of the
obstacles [12]. Simply put, the APF comprises attractive and repulsive forces that steer the robot towards
the target position, while simultaneously avoiding obstacles. The potential field is expressed
mathematically as follows:
𝑈𝑈 = 𝑈𝑈𝑎𝑎𝑎𝑎𝑎𝑎 + 𝑈𝑈𝑟𝑟𝑟𝑟𝑟𝑟 (3)
In the (3): The attractive potential field is represented by Uatt , while the repulsive potential field is
represented by Urep.
The APF methodology entails devising an abstract potential field for the robot's workspace,
comprised of the superimposition of the attractive gravitational field of the target location and the
repulsive field of obstacles. This APF can be mathematically represented as a function, and the
gravitational and repulsive forces can be defined as the negative gradients of their respective fields. By
using the spatial dynamics equation and the Lagrange equation, the force F acting on the robot due to
the APF can be obtained:
𝐹𝐹 = 𝐹𝐹𝑎𝑎𝑎𝑎𝑎𝑎 + 𝐹𝐹𝑟𝑟𝑟𝑟𝑟𝑟 (4)
In the (4): Fatt is the attraction of the target position to the robot, Frep is the repulsion of the obstacle
to the robot.

Figure 1. Force analysis of robot in APF.

168
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

Figure 1 gives a clear picture of the specific force analysis of the robot.

2.1. Gravitational field

The position of the robot in the workspace is defined as 𝑋𝑋 = (𝑥𝑥, 𝑦𝑦)𝑇𝑇 , the gravitational potential function
defined as [13]:
1 2
𝑈𝑈𝑎𝑎𝑎𝑎𝑎𝑎 = 𝑘𝑘�𝑋𝑋 − 𝑋𝑋𝑔𝑔 � (5)
2
The gravitational field that exists between the robot and the destination is mathematically represented
by Equation (5), where the constant 'k' indicates the intensity of the gravitational potential field, X
denotes the position vector of the robot, and 𝑋𝑋𝑔𝑔 signifies the target position within the potential field.
The attractive force that guides the robot towards the target is modeled by the negative gradient of the
target potential function [14]:
𝐹𝐹𝑎𝑎𝑎𝑎𝑎𝑎 = −𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑈𝑈𝑎𝑎𝑎𝑎𝑎𝑎 ) = −𝑘𝑘�𝑋𝑋 − 𝑋𝑋𝑔𝑔 � = 𝑘𝑘�𝑋𝑋𝑔𝑔 − 𝑋𝑋� (6)

2.2. Repulsion field

Khatib utilized a function called “Force Inducing an Artificial Repulsion from the Surface” as the
repulsion potential function [15]:
1 1 1 2
𝜂𝜂 � − � , 𝜌𝜌 ≤ 𝜌𝜌0
𝑈𝑈𝑟𝑟𝑟𝑟𝑟𝑟 = �2 𝜌𝜌 𝜌𝜌0 (7)
0 , 𝜌𝜌 > 𝜌𝜌0
Equation (7) defines the repulsive force field of the obstacle as 𝑈𝑈𝑟𝑟𝑟𝑟𝑟𝑟 . The potential field constant for
repulsive force, represented by η, is a positive value. The distance between the robot and an obstacle in
space position is denoted by ρ, and the parameter 𝜌𝜌0 indicates the maximum range of influence that a
single obstacle can exert.
If the distance between the robot and an obstacle is less than or equal to 𝜌𝜌0 , the repulsive potential
field will impact the robot's movement. Consequently, the repulsive force on the robot can be derived
by computing the negative gradient of the repulsive potential function. Conversely, if the distance
exceeds 𝜌𝜌0 , the repulsive potential field will have no effect on the robot's motion. [14]:
1 1 1 𝜕𝜕𝜕𝜕
𝜂𝜂 �𝜌𝜌 − � 2 , 𝜌𝜌 ≤ 𝜌𝜌0
𝐹𝐹𝑟𝑟𝑟𝑟𝑟𝑟 = −𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑈𝑈𝑟𝑟𝑟𝑟𝑟𝑟 ) = � 𝜌𝜌0 𝜌𝜌 𝜕𝜕𝜕𝜕 (8)
0 , 𝜌𝜌 > 𝜌𝜌0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑇𝑇
In the formula: = ( )
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
Of course, other functions can also be used, but it must be ensured that the functions used, and their
derivatives are continuous. The above only considers the control force of a single obstacle on the robot.
For multiple obstacles, the repulsive potential field of multiple obstacles can be obtained by
superposition of potential fields.
At this point, the total potential field U can be expressed as [16,17]:
𝑈𝑈 = 𝑈𝑈𝑎𝑎𝑎𝑎𝑎𝑎 + ∑𝑈𝑈𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 (9)
The resultant force of gravity and repulsion is:
𝐹𝐹 = 𝐹𝐹𝑎𝑎𝑎𝑎𝑎𝑎 + ∑𝐹𝐹𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 (10)

2.3. The problem of APF

The main problem of APF in generating path is prone to local minima. The local minimum point means
that the lowest point of potential energy encountered by the robot in the APF is not the global lowest
point, but a local lowest point. This situation will cause the robot to wander back and forth near the local
lowest point that cannot reach the target position, and cannot find the global lowest point or target
position. This is because, as the robot approaches the target or obstacle, the gradient in the potential field
becomes more and more gentle, and the robot is prone to local minima and cannot cross the potential
peak [18].

169
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

3. A-star algorithm
The A-star algorithm is a highly effective path planning algorithm due to its simplicity, speed, and
targeted heuristic search. It can efficiently narrow the search space and reduce problem complexity,
achieving high path search efficiency. During robot path planning, the evaluation function is used to
determine the nodes that need to be passed through at the next moment. Real-time evaluation of adjacent
nodes is required, and In the A-star algorithm, the node with the smallest evaluation function value is
chosen as the next position for the robot to move towards the target point, which allows the algorithm
to continuously approach the optimal path. The evaluation function considers both actual cost and
estimated cost, measured in Euclidean distance units, the evaluation function is as follows [19]:
f(n) = g(n) + h(n) (11)
2
g(n) = D(n − 1, n) = �(xn − xn−1 ) + (yn − yn−1 ) 2 (12)
2 2
h(n) = D(n, p) = ��xp − xn � + �yp − yn � (13)
In the (11), f(n) is the evaluation function of node n in the workspace, which represents the total cost
of the search and rescue robot at the node n in the path planning process.
In the (12), g(n) denotes the actual cost of the robot moving from the parent node of n to the node n
in the workspace.
In the (13), the function h(n) denotes the approximate cost that the robot will incur while moving
from the current node n to the destination node in the workspace. As it provides the heuristic information
required for the search process, h(n) is a vital component in the evaluation function.
The success of the A-star algorithm is directly dependent on selecting an appropriate evaluation
function, which is determined by the specific situation at hand. The key to selecting an appropriate
evaluation function lies in choosing an appropriate heuristic function, as an inappropriate one will
negatively impact the quality of the path planning. A heuristic function that more accurately estimates
the actual cost will yield better results. The estimated value of h(n) restricts the evaluation function f(n)
when the actual cost of moving from the parent node to the current node, g(n), is fixed. Therefore, nodes
closer to the target point have smaller values of h(n), resulting in smaller values of f(n) and ensuring
that the algorithm continuously moves towards the target point while searching for the shortest path.

4. Improved Artificial Potential Field

Since the environment is often unknown or partially known in practical applications, a single global
path planning cannot cope with various situations. Using a single A* algorithm, path planning can be
performed in a scene known to the environment. In the event of an unexpected change in the
environmental scene while the robot is in motion, it may be unable to evade obstacles and arrive
precisely at the target point, leading to a failed path planning. For example, some equipment distributed
in the intelligent warehouse is known, and the cargo handling trolley can plan a path based on the known
information to move the goods from the location to the destination. However, if some goods are
temporarily stacked, the global path planning cannot be accurately carried out at this time, resulting in
the cargo car unable to reach the target position. Therefore, a single A* algorithm cannot meet the actual
needs of path planning [20]. Using only the APF method for path planning may result in low efficiency
in generating paths as it is a local path planning algorithm, and lacks the ability to gain a global
understanding of the environment. In addition, it has the defects of local minima and unreachable targets,
so it will increase the amount of computation, fall into the dead cycle of path planning, and also lead to
the failure of final path planning.
Based on the above research, both the single A* algorithm and the APF method have their limitations
and are difficult to cope with special situations. Thus, to overcome the limitations of both global and
local path planning methods, a hybrid approach is proposed in this study. The proposed method
integrates the advantages of both methods, i.e., the ability to handle known environments and the ability
to adapt to unknown environments, to achieve more efficient and effective path planning for mobile
robots. In practical path planning, sensors are utilized to gather information and initialize the
environment, establishing a comprehensive map. Building on this foundation, the enhanced A*

170
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

algorithm is employed for global path planning, resulting in an optimal initial path. At the same time,
the sensor is used to continue to obtain the surrounding environment, and the IAPF method is used for
local path planning.

Figure 2. Mixed path planning flow chart.

Figure 2 provides a concrete form of the hybrid algorithm:
1. Using sensors for information collection and map construction: Using the sensors carried by the
robot for environmental perception to convert environmental information into digital maps. Through
map construction, static obstacles and dynamic obstacles around the robot can be identified and located.
2. Global path planning involves finding an optimal path using the improved A* algorithm after the
map has been constructed. This technique is primarily used for long-distance movement, such as moving
a robot from a starting point to a target point. To achieve efficient and safe path planning in the global
stage, it is crucial to consider the robot's motion capabilities and environmental constraints.
3. Local path planning: In the process of robot travel, due to environmental dynamics and sensor
errors and other reasons, there may be some local obstacles, then the robot needs to carry out local path
planning. The APF method can be used for local path planning to generate a path to avoid obstacles.
This path is usually used for detailed movement, such as the robot bypassing static obstacles or
bypassing dynamic obstacles.
4. Real-time update path: In the process of robot travel, due to factors such as environmental changes
and sensor errors, path planning needs to be updated in real time to ensure the safety and efficiency of
robot motion. Therefore, in the process of robot movement, it is necessary to continuously obtain sensor
data, update maps and path planning, and control the movement of the robot.

5. Experimental simulation
Using the planning method of the hybrid algorithm discussed in the previous section, we constructed a
simulation environment in MATLAB. The environment is a 2D space of size 25*20, with 26 randomly
placed red obstacles of radius 0.5. The robot is modeled as a circle with radius 0.2 and its starting and
ending points are marked on the map. To showcase the efficacy of the hybrid algorithm, we conducted
two sets of simulation experiments with different starting points, comparing the performance of the
hybrid algorithm with that of the single artificial potential field algorithm.

171
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

6. Experimental results

(a) (b)

(c) (d)
Figure 3. Comparison of simulation results for robot obstacle avoidance using APF and IAPF (a)
Results of robot obstacle avoidance simulation with conventional APF algorithms; (b) Result of robot
obstacle avoidance simulation with IAPF algorithms; (c)Results of robot obstacle avoidance
simulation with conventional APF algorithms after changing the start point; (d)Results of robot
obstacle avoidance simulation with conventional IAPF algorithms after changing the start point.
The experimental results on (a)(c) of Figure 3 show that the robots in the two groups of single artificial
potential field algorithms have encountered local minima. In the optimized combination algorithm, (b)(d)
of Figure 3, the robot does not fall into the local minimum point again, and changes some forward
direction, and obtains a more reasonable path. Even though the starting point was changed, the robot
was able to complete the path planning and obstacle avoidance.
The proposed hybrid path planning algorithm effectively addresses the issue of local algorithm
oscillation around obstacles, which can prevent the robot from reaching the target point. By combining
global and local path planning methods, the robot can have a comprehensive understanding of the
environment and avoid falling into local minimum values, which is a limitation of using only local path
planning. Additionally, the use of local path planning allows for a shorter planned path. The results of
the simulations show that the proposed hybrid path planning algorithm is practical and effective.

7. Conclusion
The current study proposes an innovative technique for mobile robot obstacle avoidance utilizing APF.
A-star is integrated into the development of the APF to enhance the ability to avoid local optima in the

172
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

proposed approach. Additionally, the hybridization of the two algorithms results in better planning
outcomes by complementing each other. The simulation results showcase the efficacy of the hybrid
algorithm in mitigating the issue of the robot getting stuck in local minima. However, some limitations
still exist in the traditional A* algorithm combined with the APF method. For instance, it is unable to
detect local minima of random multiple individual obstacles, which calls for further optimization of the
algorithm model. Furthermore, the investigation of dynamic and sudden obstacles will be a future
research direction for this study.

References
[1] Mansur M T, Dissanayake R L, and Gamage M W M. 2006 Obstacle avoidance in mobile robots
using randomized escape strategy. In Proc. IEEE International Conference on Robotics and
Automation (ICRA), pp: 1109-1114.
[2] Gao Y, Wang X, and Song Z. 2017 A novel potential field-based path planning method using
heuristic search for mobile robots. Journal of Intelligent and Robotic Systems, 90(1), pp: 19-
34.
[3] Cho Y J, Oh S Y, and Kim K S. 2002 Path planning for mobile robots using a modified harmonic
potential field method. IEEE Transactions on Robotics and Automation, 18(5), pp: 710-723.
[4] Xie J, and Inami M. 2014 Obstacle avoidance in complex environments by a simulated annealing
algorithm with a human-like learning capability. In Proc. IEEE International Conference on
Robotics and Biomimetics (ROBIO), pp: 1409-1414.
[5] Roy S, and Roy A B. 2014 Path planning of mobile robot using potential field and semi-
landmarking techniques. International Journal of Advanced Research in Computer and
Communication Engineering, 3(8), pp: 7289-7293.
[6] Alzugaray S, Unzueta L, and Alonso J M. 2013 A fuzzy artificial potential field for autonomous
mobile robot navigation. Robotics and Autonomous Systems, 61(2), pp: 121-134.
[7] Kucuk G B, Temeltas H, and Ozturk S. 2010 Extremal optimization-based artificial potential field
method for obstacle avoidance of mobile robots. Robotics and Autonomous Systems, 58(2), pp:
179-186.
[8] Mukhopadhyay S C, Acharjee S and Sanyal S. 2017 A wall-following mobile robot using artificial
potential field and modified fuzzy logic controller. International Journal of Advanced Robotic
Systems, 14(6), pp: 1-13.
[9] Yu L, Li S, and Fu X. 2018 An obstacle avoidance method for mobile robots based on virtual
flow field and modified artificial potential field. In Proc. IEEE International Conference on
Mechatronics and Automation (ICMA), pp: 1471-1476.
[10] Sha C, Guangming M, and Xinqing Wu. 2019 Path planning for mobile robots based on improved
artificial potential field method. Computer Applications, 39(5), pp: 1385-1390.
[11] Shuqin Z, lin Y, and Yang L. 2018 A multi-robot path planning algorithm based on the artificial
potential field method. Computer Engineering and Applications, 54(5), pp: 183-188.
[12] Jia J, and Wang J. 2019 Robot path planning algorithm based on improved artificial potential field
method. Proceedings of the 2019 9th International Conference on Measuring Technology and
Mechatronics Automation, pp: 534-538.
[13] Yanhui L, Guangming, M., and Guang, C. 2019 Path planning for mobile robots based on
dynamic artificial potential field method. Computer Applications, 45(2), pp: 77-82.
[14] Guo C, and Wu H. 2020 Robot path planning algorithm based on a hybrid artificial potential field
and genetic algorithm. Soft Computing, 24(3), pp: 1833-1847.
[15] Yang C, and Li M. 2018 A new artificial potential field algorithm for multi-robot path planning.
Journal of Intelligent and Fuzzy Systems, 34(1), pp: 33-40.
[16] Yufeim W, Xin L, and Donglin, Zhao. 2018 Research on UAV path planning based on artificial
potential field method. Computer Engineering and Design, 39(6), pp: 1487-1492.

173
 Proceedings of the 2023 International Conference on Mechatronics and Smart Systems
DOI: 10.54254/2755-2721/10/20230170

[17] Lin Y, and Su, Y. 2018 A hybrid potential field and particle swarm optimization algorithm for
mobile robot pathplanning. International Journal of Advanced Robotic Systems, 15(4),
1729881418781316.
[18] Li H, Li L, and Li D. 2020 Path planning for mobile robots based on improved artificial potential
field algorithm. International Journal of Advanced Robotic Systems, 17(1),
1729881420903005.
[19] Zhibin C, Wei W, and Weitao L. 2018 Robot path planning based on multi-objective artificial
potential field method. Computer Applications and Software, 35(3), pp: 186-191.
[20] Sun H, Zhu H, and Zhang Y. 2018 A path planning method for mobile robots based on improved
artificial potential field algorithm. Proceedings of the 2018 3rd International Conference on
Robotics and Automation Engineering, pp: 128-132.

174