Rolling-horizon genetic algorithm for adaptive path planning in hazardous environments

초록

Effective path planning in hazardous environments, such as those involving toxic gas leaks or radiation exposure, require algorithms that can adapt to dynamic and uncertain conditions in real time. This paper proposes a Rolling-Horizon Genetic Algorithm (RHGA) designed to address the limitations of both deterministic and conventional evolutionary methods in such settings. RHGA enhances adaptability and scalability by incrementally optimizing and fixing short path segments, rather than evolving entire trajectories at once. It incorporates a hybrid adaptive selection mechanism, diversity-preserving crossover, and fixed-path anchoring to guide convergence. Additionally, the population is partitioned into subgroups targeting different objectives, including exposure minimization, distance reduction, and balanced tradeoffs, and optimized in parallel to support multi-criteria planning. Comprehensive experiments in grid-based environments with dynamic gas diffusion, hazard injection, and negative-cost edges demonstrate that RHGA consistently outperforms various baseline algorithms in both adaptability and computational efficiency. Theoretical analysis and empirical results confirm that RHGA maintains responsiveness under evolving risks while producing diverse, high-quality escape paths suitable for real-time deployment.

키워드