Neural Controller for a 5-Link Planar Biped Robot

The canonical problems in control of the biped robots arise from underactuation, impulsive nature of the impact with the environment and existence of the many degrees of freedom in their mechanism. Since biped walkers have fewer actuators than degrees of freedom, they are underactuated mechanical systems. In this paper according to the humans and animals locomotion algorithms, the stability of an underactuated biped walker with point feet is done by central pattern generator and feedback networks. For tuning the parameters of the CPG network, the control problem is defined as an optimization problem. This optimization problem is solved by using of genetic algorithm. Also a new feedback structure is proposed for the biped walker. This feedback network can overcome the difficulty of the coordination of the knee and the hip joints. Also PI controllers are used as servo controllers. Finally, the effectiveness of the proposed methods is confirmed by simulation results.