Stepping of biped robots over large obstacles using NMPC controller

Abstract

One of the main challenges for biped robots is to step over large obstacles during walking. In this paper, a control method is proposed for walking and stepping over large obstacles based on the Nonlinear Model Predictive Control (NMPC) method. One of the main advantages of the proposed method is that it is trajectory free, which gives the robot the ability to step over any feasible obstacle. Moreover, the NMPC guarantees dynamic stability during walking and crossing over the target. In addition, a multilayer perceptron neural network is employed for identification of the dynamic model of the robot. In this way, the proposed method can cope with uncertainties in the robot model. Simulation results show good performance of the proposed method applied to a 173 cm robot stepping over a 40×15cm obstacle dynamically in the sagittal plane while maintaining a safety clearance from it, without any need for reference trajectory.