Practical Aspects of Artificial Potential Field Implementation Based on 2D LiDAR Scans in an Autonomous Mobile Robot

The Artificial Potential Field (APF) is a popular local path planner for mobile robots. Its many modifications assume a given set of obstacles, although on a real robot, this set must be produced by point cloud segmentation of a light detection and ranging (LiDAR) scan. It is worth noting that this stage is not always specified in the literature on APF-based methods. Therefore, the selected segmentation methods from three families, i.e., breakpoint, density-based, and geometric, are compared in this paper. Moreover, rules for generating the velocity command based on the gradient of the potential function presented in the literature are compared on the running system to present the two most important aspects of APF implementation on a real mobile robot. The results make it possible to determine the most useful segmentation method, taking into account the computational requirements and the workspace of the mobile robot. The obtained results show that the segmentation method does not provide any improvement for simple, well-shaped obstacles. However, obstacles with a small gap between them can be separated as two obstacles or as only one, which causes changes in the potential field and slightly decreases the smoothness of the mobile robot path. In such a scene, two of the five methods took 0.38–0.51 s longer and moved 13–14% less smoothly than breakpoint detection with the range-adaptive threshold. Regardless of the path-quality indices, the computational effort of segmentation methods differs significantly. The simplest segmentation method requires about 250 times less computation than the most expensive one. The rule for control signal generation based on the gradient of the potential field has a significant impact on the path-quality indices. The main observation is that smoother movement and a more repeatable distance to the obstacle can be achieved using the velocity profiling based on the orientation error, but the mobile robot will need more time to reach the goal position. It should be emphasized that the computational effort of the APF and the control rules is significantly lower than that of the segmentation. Therefore, the main computational cost of APF implementations lies in point cloud segmentation into useful data for potential field calculation.

Authors

Institutions

Publication Details

Journal
Robotics
Published
2026-11-19
DOI
https://doi.org/10.3390/robotics15110244
Primary Topic
Robotic Path Planning Algorithms
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Practical Aspects of Artificial Potential Field Implementation Based on 2D LiDAR Scans in an Autonomous Mobile Robot

Rafał Szczepański, Kamil Przeczewski
Robotics
Robotic Path Planning Algorithms
article

Practical Aspects of Artificial Potential Field Implementation Based on 2D LiDAR Scans in an Autonomous Mobile Robot

Rafał Szczepański, Kamil Przeczewski
article en

Abstract

The Artificial Potential Field (APF) is a popular local path planner for mobile robots. Its many modifications assume a given set of obstacles, although on a real robot, this set must be produced by point cloud segmentation of a light detection and ranging (LiDAR) scan. It is worth noting that this stage is not always specified in the literature on APF-based methods. Therefore, the selected segmentation methods from three families, i.e., breakpoint, density-based, and geometric, are compared in this paper. Moreover, rules for generating the velocity command based on the gradient of the potential function presented in the literature are compared on the running system to present the two most important aspects of APF implementation on a real mobile robot. The results make it possible to determine the most useful segmentation method, taking into account the computational requirements and the workspace of the mobile robot. The obtained results show that the segmentation method does not provide any improvement for simple, well-shaped obstacles. However, obstacles with a small gap between them can be separated as two obstacles or as only one, which causes changes in the potential field and slightly decreases the smoothness of the mobile robot path. In such a scene, two of the five methods took 0.38–0.51 s longer and moved 13–14% less smoothly than breakpoint detection with the range-adaptive threshold. Regardless of the path-quality indices, the computational effort of segmentation methods differs significantly. The simplest segmentation method requires about 250 times less computation than the most expensive one. The rule for control signal generation based on the gradient of the potential field has a significant impact on the path-quality indices. The main observation is that smoother movement and a more repeatable distance to the obstacle can be achieved using the velocity profiling based on the orientation error, but the mobile robot will need more time to reach the goal position. It should be emphasized that the computational effort of the APF and the control rules is significantly lower than that of the segmentation. Therefore, the main computational cost of APF implementations lies in point cloud segmentation into useful data for potential field calculation.

RoboticsVol. 15(11)
Nicolaus Copernicus University (PL)
Robotic Path Planning Algorithms
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon