FTDRL-HGSO: an efficient terrain-adaptive DRL-driven gait and path control framework for hexapod robots

This model builds on Fusion terrain aware Deep Reinforcement Learning (FTDRL) methodology by means of introducing FTDRL-HGSO (Henry Gas Solubility Optimization) methodology, which is employed for adaptive gait control & path optimization for hexapod robots operating in a varied terrain condition. The suggested scheme integrates multi-terrain analysis, dynamic gait adaptation, and energy efficient locomotion through hybrid learning-optimization model. On coupling a DRL policy network with HGSO metaheuristic, the model improves exploration-exploitation balance and thus accelerates convergence towards optimal locomotion strategies. The terrain analysis module thus extracts the elevation and surface characteristics, that are then processed for adapting gait phase and control parameters dynamically. Simulation studies conducted in MATLAB-Simulink under varying terrain conditions like sand, clay and rock shows that suggested FTDRL-HGSO model attains enhanced convergence, improved stability, and decreased consumption of energy on comparing benchmark models like PSO, TGPSO, GWO, and NGO. Experimental outcome shows that FTDRL-HGSO model attains better performance on various metrics like stability index, speed indicator, COT index and fitness function convergence analysis on varied terrain conditions like rock, sand, and clay. The proposed model attains low fitness function value of 0.165 on clay terrain and higher stability index of 0.69 outperforming other existing models in both convergence efficiency and terrain adaptability.