Home // SIMUL 2025, The Seventeenth International Conference on Advances in System Modeling and Simulation // View article

Generalizable Spatiotemporal Reinforcement Learning Model for Maritime Search Path Planning

Authors:
Pengcheng Yang
Yingying Gao
Jing Xu
Qingqing Yang

Keywords: Keywords-maritime search and rescue; reinforcement learning; generalization ability; path planning

Abstract:
Maritime search path planning is critical for enhancing search efficiency and seizing the golden rescue time in maritime search and rescue operations. To address the insufficient generalization of existing methods, this paper presents a spatiotemporally enhanced Reinforcement Learning (RL) model. By simulating the target's probability distribution via a mixed Gaussian distribution and incorporating a Long Short-Term Memory (LSTM) network into the Proximal Policy Optimization (PPO) approach, the model's ability to extract spatiotemporal features is enhanced. Furthermore, a threshold-based scenario-switching mechanism is designed to boost training stability. Experimental results demonstrate the model's exceptional generalization and significantly improved solution quality on both training and test sets.

Pages: 79 to 81

Copyright: Copyright (c) IARIA, 2025

Publication date: September 28, 2025

Published in: conference

ISSN: 2308-4537

ISBN: 978-1-68558-300-2

Location: Lisbon, Portugal

Dates: from September 28, 2025 to October 2, 2025