Self-improving biped locomotion

An approach addressing biped locomotion is here introduced. Central Pattern Generators (CPGs) and Dynamic Movement Primitives (DMPs) were combined to easily produce complex trajectories for the joints of a simulated DARwIn-OP. Policy Learning by Weighting Exploration with the Returns (PoWER) was implemented to improve the robot's locomotion through variation of the DMP's parameters. Maximization of the DARwIn-OP's frontal velocity was addressed and results show a velocity improvement of 213%. The results are very promising and demonstrate the approach's flexibility at generating new trajectories for locomotion.