FAHP-guided LSSVM-PSO for stability modeling and parameter identification of underground heavy-load robots

This paper proposes a novel stability modeling and parameter identification framework that integrates Fuzzy Analytic Hierarchy Process (FAHP) with a hybrid Least Squares Support Vector Machine and Particle Swarm Optimization (LSSVM-PSO) algorithm. First, FAHP quantifies the influence weights of four coupling dimensions—environmental-mechanical, mechanical-body, perception-environment, and electro-hydraulic-mechanical—identifying track-soil compaction (weight 0.140) and electro-hydraulic response (weight 0.112) as the most critical factors. Guided by these weights, an LSSVM-PSO algorithm is developed to accurately calibrate 12 key uncertain parameters (e.g., soil deformation modulus, electro-hydraulic response coefficients) using multi-source sensor data. Laboratory experiments and underground field tests demonstrate that the optimized model reduces the comprehensive prediction error from 9.2% to 3.9%, achieves a 61.6% reduction in Zero-Moment Point (ZMP) prediction error, and lowers attitude angle RMSE by over 60%. In field tests under harsh mining conditions, the model achieves 90% accuracy in predicting stability boundary violations and improves lifting operation precision by 16% through real-time decision support. This study provides a systematic, data-driven methodology for high-fidelity stability modeling of heavy-load robots in complex environments. It offers both theoretical insights into multi-source coupling mechanisms and practical tools for stability monitoring and predictive control in underground mining applications.