Energy Efficiency and Speed Optimization by Squad-Unit Genetic Algorithm for Bipedal Walking

In this paper, a method of motion planning and trajectory optimization is proposed to decrease energy consumption and improve speed performance for bipedal walking. For a long time, researches of bipedal walking have been concentrated on single supporting phase, in which dynamic model such as inverted pendulum model (IPM) is used. Originally, in this paper the trajectory of the center of mass (COM) is generated separately: trajectory of COM is first planned in double supporting phase (DSP), then the trajectory of COM in single supporting phase (SSP) is generated consequently based on inverted pendulum model (IPM). Trajectory of COM in DSP, derivating SSP trajectory is optimized to achieve minimum energy consumption and better speed performance for bipedal walking. The evaluation of energy consumption and speed is illustrated as total cost of transporting (TCOT). To optimize a three dimensional space trajectory in DSP, modified genetic algorithm (GA), squad-unit generic algorithm is conducted. In this method, the unit for evolution contains three members sharing the same fitness value and has the advantage over conventional that can only represent one-dimensional input variable. The effect of optimization of three dimensional trajectories is verified by simulation.