Real-time Gait Generation for Humanoid Robot Based on Fuzzy Neural Networks

Abstract

Gait parameters of humanoid robot are optimized by introducing three energy consumption indexes, calculation formulas of driving torque for each joint of humanoid robot are derived based on Lagrange dynamics equation, mathematic models for gait optimization are established. A set of optimal solutions for gait parameters are obtained utilizing Matlab as simulation and optimization tool. The minimum consumed energy gait, which similar with human motion, are used to teach the fuzzy neural network(FNN), after supervised learning, the FNN can quickly generate the humanoid robot gait parameters. A new approach for real-time gait planning of humanoid robot during walking is proposed based on FNN, ZMP criteria, B-spline interpolation and inverse displacement analysis model. Simulation results demonstrate feasibility and effectiveness of the proposed real-time planning method. Numeric examples for gait optimization and real-time gait planning of humanoid robot are given for illustration.