GAA-DFQ: A Dual-Layer Learning Model for Robot Path Planning in Dynamic Environments Integrating Genetic Algorithms, DWA, and Q-Learning

This paper presents a dual-optimization learning model combining genetic algorithms for global path planning with local obstacle avoidance algorithms for robot navigation in dynamic environments. The model integrates <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A^{*}$</tex> for global path planning, and the Dynamic Window Approach (DWA), fuzzy control, and Q-learning for real-time obstacle avoidance. An improved multi-objective genetic algorithm is used to optimize path length, safety, and smoothness. Experimental results demonstrate that the GAA-DFQ algorithm outperforms traditional <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{A}^{*}$</tex> and GAA in path planning, showing the lowest collision rate (2.5 %), the shortest path length (181 units), and the fastest computation time (85 seconds), proving its efficiency and stability in complex environments.