Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization

Hongyu Cao; Weixiang Gan

doi:10.47297/taposatWSP2633-456908.20260708

PDF
导出
分享
收藏
专辑

Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization

7卷第8期

浏览量: 0 Downloads: 0

Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization
Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization
1.Langfang Zhenguigu Technology Co., Ltd., Langfang, Hebei, 065000, China
2.Lincoln University College, Petaling Jaya, Selangor Darul Ehsan, 47301, Malaysia
2026年7卷第8期页码：41-45
纸质出版：2026-08-30
DOI：10.47297/taposatWSP2633-456908.20260708
Accepted：

Scan QR Code

Hongyu Cao & Weixiang Gan. (2026). Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization. 科技理论与实践（英文版）, 7(8), 41-45.

Cao Hongyu,Gan Weixiang.Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization[J].Theory and Practice of Science and Technology,2026,07(08):41-45.
Hongyu Cao & Weixiang Gan. (2026). Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization. 科技理论与实践（英文版）, 7(8), 41-45. DOI： 10.47297/taposatWSP2633-456908.20260708.

Cao Hongyu,Gan Weixiang.Machine Learning-Driven Optimization of Mobile Robot Obstacle Avoidance: DQN-Based Simulation Data Analysis and Visualization[J].Theory and Practice of Science and Technology,2026,07(08):41-45. DOI： 10.47297/taposatWSP2633-456908.20260708.

摘要

Abstract

This paper models mobile robot obstacle avoidance as a Markov decision process and trains a deep Q-network (DQN) on a 29-dimensional state of 24 lidar ranges and the relative goal pose. Across 600 random scenarios at three obstacle densities

the DQN reaches 90.5% success against 83.7% for the artificial potential field method (APF) and 84.0% for the dynamic window approach (DWA) under identical sensing (χ² = 11.85 and 10.82

p ≤ 0.001); it never times out

and its median travel time of 23.4 s is two thirds of DWA's 35.0 s. A reward ablation prices each term: the sparse reward yields 0% success

progress shaping 63.5%

a proximity penalty 92.0%

lifting the median margin from 0.447 m to 0.755 m. Seven figures trace each difference to its mechanism

and 143 of 151 collisions (94.7%) involve dynamic obstacles.

关键词

Keywords

references

Chen, Y. F., Liu, M., Everett, M., & How, J. P. (2017). Decentralized non-communicating multiagent collision avoidance with deep reinforcement learning. 2017 IEEE ICRA , 285-292.

Everett, M., Chen, Y. F., & How, J. P. (2018). Motion planning among dynamic, decision-making agents with deep reinforcement learning. 2018 IEEE/RSJ IROS , 3052-3059.

Fox, D., Burgard, W., & Thrun, S. (1997). The dynamic window approach to collision avoidance. IEEE Robotics & Automation Magazine , 4(1), 23-33.

Hart, P. E., Nilsson, N. J., & Raphael, B. (1968). A formal basis for the heuristic determination of minimum cost paths. IEEE Transactions on Systems Science and Cybernetics , 4(2), 100-107.

Khatib, O. (1986). Real-time obstacle avoidance for manipulators and mobile robots. The International Journal of Robotics Research , 5(1), 90-98.

文章被引用时，请邮件提醒。

Submit

Research on Path Planning Algorithm Combining Deep Reinforcement Learning and Theta* Algorithm

Multi-Agent Deep Reinforcement Learning in UAV Swarm Formation Control: A Review and Future Challenges

Visualization Design and Implementation of Enterprise Expense Data Based on the Whole Value Chain

A Review of the Applications and Innovative Practices of Artificial Intelligence Algorithms in Gaming

Research on Perception and Decision-Making Methods for Autonomous Driving

Related Author

Haoyang Qin School of Software, Shandong University

Wenchao He School of Intelligent Engineering, Yanching Institute of Technology

Shuang Wu School of Management, Liaoning University of International Business and Economics, Information Management and Information Systems

Yihan Liu Anhui University

Qinwen Qiu School of Management and engineering, Nanjing University, Industrial Intelligence and System Integration

Jundan Wang University of Science and Technology Liaoning

Aobo Zhang University of Science and Technology Liaoning

Jikai Xin University of Science and Technology Liaoning

Related Institution

School of Software, Shandong University

School of Intelligent Engineering, Yanching Institute of Technology

School of Management, Liaoning University of International Business and Economics, Information Management and Information Systems

Anhui University

School of Management and engineering, Nanjing University, Industrial Intelligence and System Integration

⁰