Improving Quadrupedal Locomotion on Granular Material Using Genetic Algorithm

This paper describes process of generating a stable gait for a quadruped robot on granular material such as sand. To achieve this goal a simulated environment was created, to model the kinematics and dynamics of the robot. Genetic algorithm was used as a gait generation technique. The simulation model of Laikago was built in MATLAB and Simscape. The contact model between the foot and ground, was based on existing results from sand penetration testing. The genetic algorithms optimised the joint space trajectory, to evaluate of the gait performance on granular terrain. The proposed method required to first create a simulation constrained to planar motion to simplify the problem. The 2D gait was then used as the initial guess for the optimisation in 3D simulation and allowed for creation of an acceptable gait. The final gait has demonstrated the effectiveness of the proposed approach for quadruped robot walking on granular terrain.