A real time path generator for fast motion of a biped carrying different payloads

Stability and performance are two main issues in motion of bipeds. To ensure stability, motion of a biped needs to follow specific pattern to comply with certain stability criterion such as zero moment point. Actuator torques limitation is also a serious restriction on motion path. Indeed, the velocity of the robot during each step is bounded with robot´s physical limitations. The motion path of the robot is designed considering environmental conditions and robot´s tasks. Therefore, online trajectory planning is one of the most important problems in this field. In this paper a neural network based algorithm for stable path generation is introduced which can be used online to generate path for a biped carrying different payloads. The neural network is trained based on optimum paths generated by a genetic algorithm for different values of payload mass, while considering actuator limits as optimization constraints.