Adaptive Obstacle Avoidance Algorithm for Robotic fish integrating Recurrent Proximal Policy Optimization and Improved Dynamic Window Approach
Abstract
underwater aquaculture environments, the traditional Dynamic Window Approach (DWA) with a fixed weight mechanism exhibits poor
balance among obstacle avoidance safety, path smoothness, and energy efficiency. To this end, this paper proposes an adaptive obstacle
avoidance algorithm for robotic fish based on Recurrent Proximal Policy Optimization and Dynamic Window Approach (RPPO-DWA).
First, an improved DWA evaluation function adapted to the underwater three-dimensional environment is constructed by introducing
height error, dynamic collision risk, and energy consumption constraints to enhance decision quality. Second, the weight adjustment
process is modeled as a Partially Observable Markov Decision Process (POMDP), utilizing a Gated Recurrent Unit (GRU) network
to extract temporal environmental features and solving for the optimal weight vector online via RPPO. Finally, the Generalized StateDependent Exploration (gSDE) mechanism is introduced to suppress path oscillation. Simulation results demonstrate that, compared
with the fixed-weight DWA, the proposed algorithm significantly improves the obstacle avoidance success rate and navigation efficiency,
effectively achieving predictive obstacle avoidance and synergistic optimization of energy efficiency for robotic fish under conditions of
time delay and dynamic disturbances.
Keywords
Full Text:
PDFReferences
[1] Zhu Yijue, Du Xiujuan. Review and Prospect of Underwater Environmental Monitoring System [J]. Computer Engineering and Applications, 2023, 59(10): 65-74.
[2] Yang Meihui, Hao Zixin, Li Bin. Research Analysis and Review of Robot Path Planning Algorithms [J]. Journal of Qilu University of
Technology, 2024, 38(05): 37-46.
[3] Cheng Mingyu, Liu Ting, Yang Fang, et al. Design and Application of Biomimetic Robotic Fish [J]. Science & Technology Information,
2023, 21(14): 52-55.
[4] Cheng Xiaohui. Research on Three-dimensional Observation Path Planning Method for Mobile Platform Based on Proximal Policy Optimization [D]. Harbin Engineering University, 2023.
[5] Chen Jingyi. Research on Mobile Robot Path Planning Based on Improved Dynamic Window Approach [J]. Internet of Things Technologies, 2024, 14(03): 102-107.
[6] Wu Qiong, Gong Zaiwu, Zhang Yongchui. Research on 3D Obstacle Avoidance for UUV Based on Fusion of VO and DWA [J]. Journal
of Nanjing University of Information Science and Technology, 1-13 [2026-01-14].
[7] Teng Fei, Wang Yingchun, Yao Yonghui, et al. UAV Dynamic Obstacle Avoidance Planning Based on Deep Reinforcement Learning [J].
Journal of Beijing University of Aeronautics and Astronautics, 1-13 [2026-01-14].
[8] Mei Jie, Chen Zhe, Qin Jiarui. Path Planning Control of Robotic Fish Based on Deep Reinforcement Learning [J]. Hoisting and Conveying Machinery, 2025, (22): 34-41.
[9] Jia Keyi. Research on 3D Path Planning Algorithm for Agents Based on Improved PPO [D]. Northeast Forestry University, 2025.
[10] Huang Cikun. Research on Path Planning of Unmanned Vehicle Based on Dynamic Window Approach with Deep Reinforcement Learning [D]. Fuzhou University, 2023.
DOI: http://dx.doi.org/10.70711/aitr.v3i7.8867
Refbacks
- There are currently no refbacks.
