Evolution of Biped Walking Using Physical Simulation

Controlling a biped robot with a high degree of freedom to achieve stable movement patterns is still an open and complex problem. Thus, the development of control mechanisms for biped walking have become an important field of research. With growing calculation power of computer hardware, high resolution realtime simulation of such robot models in effect becomes all the more applicable. This thesis presents a physical simulation of a 19 degrees of freedom real biped robot model and demonstrates its application for exploring biped motion control techniques. It introduces two different approaches of biped motion generation, which specify target angle functions for all driven joints. Both approaches were developed and implemented within the physical simulation of the biped robot. Whilst the motion parameters were identified and optimized using evolutionary algorithms, several walking patterns could be generated, allowing the simulated robot for covering finally more than 10m within 25s.