Quasi optimal gait for a biped robot using genetic algorithm

The context of our research is the study of legged robots. In order to realize autonomous legged robots, the energy consumption during a step should be minimized. The method proposed is based on the expansion in Fourier's series of the joint trajectories. Thus, the search of optimal trajectories is converted into a nonlinear programming problem. The convergence cannot be guaranteed if a constraint (the floor) is introduced, since the cost function becomes highly nonlinear and multi modal. The incorporation of this constraint in terms of a simplified model for the floor and the contact force significantly improve the algorithm's convergence.