Research Of Multi-objective Gait Planning For A Biped Humanoid Robot

Walking control of biped humanoid robot has been a hotpot in robotics research field. Biped humanoid robot body structure is so complex that it is very difficult to control walking flexibility and stability of the robot. The kinematics model is usually an important prerequisite for walking. In this thesis, implementation of forward kinematics was described first in detail according to Robonova-1 specific robot model.This thesis decomposes gait into periodic movements of three independent phases according to the specialty of the robot walk. And only need to set the key pose of each movement, then using cubic spline interpolation to generate all other continuous middle poses in a walk. In this thesis, each motion was planned in detail by stages, planning out the trajectories of the ankle and hip joint, and design the motion trajectory by parameters , once the value of these gait parameters known, we will get the corresponding trajectory. So, how to set the appropriate value of gait parameters is a research problem.This thesis considers use optimization algorithm to get the appropriate value of gait parameters. Current gait optimization algorithm only achieved a single objective optimization. Because a single target optimality doesn't mean the whole optimality. In this paper, according to the diversity of robot gait objective, taking the biped robot gait optimization as multi-objective optimization problem, and constructing three objective evaluation functions about stability, power consumption and walking speed. Because of these three objectives are conflictive each other in optimization direction, we can't simultaneously obtain their optimal solution, we need to use methods which can compromise on the three objectives. As the direct weighted summation of fitness functions can not handle commendably the problem about multiple conflicting objectives, this paper uses the current relatively efficient multi-objective evolutionary algorithm SPEA2 (Strength Pareto Evolutionary Algorithm2) to solve multi-objective gait parameters optimization problem. Finally, Planning out optimization gaits meeting the three conflicting objectives. Experiment shows that the effectiveness of the algorithm, and has good application in real robot.