Research On Free Transition Control Of Hexapod Robot Wave Gait Based On Machine Learning

Hexapod robot is applied to various unstructured operating environments with its stable walking characteristics,which can replace human activities.In order to make the multi-legged robot adapt to various non-structural operating environment,the problem of gait transition and its stability has become a difficulty research hotspot.Based on six-legged robots,the stability of the Hexapod robot gait margin and machine learning-based gait transition control algorithm are given in this paper.Simulation and physical experiments are presented to verify the performance of the proposed gait transition method.In this paper,four fluctuating gait diagrams are obtained according to different land occupation coefficients.The corresponding mathematical description forms can be obtained by analyzing the fluctuating gait diagrams,respectively.And,the motion space of the robot system is determined by solving the real-time stability margin of the system.Combining the position of one foot in the discrete form and the given corresponding foot position status with constraints,we can obtain the next state selection algorithm.Based on this algorithm,the reinforcement learning algorithm of gait robot and the analysis of the gait transition process is proposed.Based on the learning algorithm,the transition process is given,then,the trajectory planning of foot position is carried out.The leg kinematics model and the solution result of the Hexapod robot are given on the base of the kinematics analysis for the single leg and the system,and the added constraint to the legs,and then,the trajectory planning of foot position is presented.Based on the simulation environment of PRO/E and ADAMS platform,the reinforcement learning based fluctuation gait transition with 1/2,2/3,5/6 occupancy coefficient is simulated to verify the feasibility and effectiveness of the proposed algorithm.Based on the physical platform,the reinforcement learning algorithm of wave gait transition process is verified in this paper.The experimental results provided in this thesis demonstrate that the reinforcement learning algorithm can achieve the gait transition efficiency of wave gait and ensure the stability of the robot in the transition process.