A Multi-Objective Path Planning Framework in Off-Road Environments Based on Deep Reinforcement Learning
IEEE Transactions on Intelligent Transportation Systems · 2025-11 · 26 (11) · 19163-19181
DOI: 10.1109/tits.2025.3589586 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Intelligent Transportation Systems, 2025-11. Verfasst von Lilu Zhu und 3 weiteren Autor:innen. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 70 Literaturangaben. Gefördert von: Frontier Technologies R&D Program of Jiangsu.
Abstract (Originalsprache)
Efficient and safe path planning is a key navigation requirement for intelligent vehicles, particularly in off-road environments characterized by complex terrain. Traditional methods focus mostly on minimizing travel time or path length, neglecting the potential for vehicle failure due to complex terrain and dynamic conditions. This limitation is further exacerbated under varying meteorological influences. Here, we propose a heuristic multiobjective path planning (AC-MOPP) framework based on deep reinforcement learning (DRL). This framework overcomes the shortcomings of traditional methods by incorporating multiple objectives and accounting for meteorological variations while reducing learning costs through DRL algorithms. Specifically, we define a traversability map environment, an RL agent, driving actions, and path evaluation methods to construct the path planning model. Using the actor–critic algorithm, we integrate meteorology-driven heuristic rules and a prioritized experience replay method to assist the agent in adjusting its path, accelerating the convergence process, and reducing its learning costs. Finally, comparative experiments are conducted, and the results demonstrate that the AC-MOPP outperforms the standard DRL and state-of-the-art non-DRL algorithms in terms of planning efficiency and convergence stability. Moreover, AC-MOPP achieves better navigation performance than do commercial platforms such as Google Maps, Bing Maps, and OpenStreetMap. This work offers a robust navigation solution for intelligent vehicles under various off-road conditions.
Förderung
- Frontier Technologies R&D Program of Jiangsu
Zitieren
APA
Zhu, L., Huang, K., Wang, Y., & Hu, Y. (2025). A Multi-Objective Path Planning Framework in Off-Road Environments Based on Deep Reinforcement Learning. IEEE Transactions on Intelligent Transportation Systems, 26(11), 19163-19181. https://doi.org/10.1109/tits.2025.3589586
Vancouver
Zhu L, Huang K, Wang Y, Hu Y. A Multi-Objective Path Planning Framework in Off-Road Environments Based on Deep Reinforcement Learning. IEEE Transactions on Intelligent Transportation Systems. 2025;26(11):19163-19181. doi:10.1109/tits.2025.3589586
BibTeX
@article{Zhu2025,
author = {Zhu, Lilu and Huang, Kai and Wang, Yang and Hu, Yanfeng},
title = {A Multi-Objective Path Planning Framework in Off-Road Environments Based on Deep Reinforcement Learning},
journal = {IEEE Transactions on Intelligent Transportation Systems},
year = {2025},
volume = {26},
number = {11},
pages = {19163--19181},
doi = {10.1109/tits.2025.3589586}
}
EUVIMED