Action-Curiosity-Based Deep Reinforcement Learning Algorithm for Path Planning in a Nondeterministic Environment
Intelligent Computing · 2025-01 · 4
DOI: 10.34133/icomputing.0140 ↗Kurz gefasst
Fachartikel in der Zeitschrift Intelligent Computing, online erschienen am 03.06.2025. Verfasst von Junxiao Xue und 2 weiteren Autor:innen. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 42 Literaturangaben.
Abstract (Originalsprache)
In the field of path planning, the efficiency and effectiveness of deep reinforcement learning (DRL) methods are often constrained by the algorithms’ exploration capabilities, particularly in dynamic and nondeterministic environments. This paper introduces a novel DRL optimization approach predicated on an action curiosity mechanism, designed to enhance both performance and efficiency in uncertain settings. By incentivizing agents to explore their surroundings more effectively, the action curiosity module amplifies learning efficiency and curtails training duration. The method’s adaptability and stability in intricate or dynamic scenarios are augmented through a dynamically adjusted reward mechanism. To mitigate the issue of strategy degradation stemming from excessive exploration, we incorporate a cosine annealing strategy that finesses parameter adjustments in real time. Extensive experimentation reveals that our enhanced algorithm outperforms conventional methods markedly in terms of success rate and average reward, among other metrics. These experimental outcomes corroborate the proposed method’s robustness and efficacy, laying a solid groundwork for efficient adaptive autonomous path planning within complex and nondeterministic environments.
Zitieren
APA
Xue, J., Chen, J., & Zhang, S. (2025). Action-Curiosity-Based Deep Reinforcement Learning Algorithm for Path Planning in a Nondeterministic Environment. Intelligent Computing, 4. https://doi.org/10.34133/icomputing.0140
Vancouver
Xue J, Chen J, Zhang S. Action-Curiosity-Based Deep Reinforcement Learning Algorithm for Path Planning in a Nondeterministic Environment. Intelligent Computing. 2025;4. doi:10.34133/icomputing.0140
BibTeX
@article{Xue2025,
author = {Xue, Junxiao and Chen, Jinpu and Zhang, Shiwen},
title = {Action-Curiosity-Based Deep Reinforcement Learning Algorithm for Path Planning in a Nondeterministic Environment},
journal = {Intelligent Computing},
year = {2025},
volume = {4},
doi = {10.34133/icomputing.0140}
}
EUVIMED