Real-Time Control of Full Actuated Biped Robot Based on Nonlinear Model Predictive Control

摘要

A trajectory free walking control scheme was proposed for actuated biped robot with the NMPC method in order to carry out real-time gait programming. The basic feature in the proposed strategy is to use iterative on-line optimization approach to compute decision variables in predictive horizon. One full stride is split into two phases which are Single Support Phase(SSP) and Double Support Phase(DSP). Correspondingly two different dynamic optimization problems are formed with seriously chosen virtual constraints. Parameterization method and penalty function method help to convert original dynamic optimization problem to static finite dimensional one which can be solved by SQP algorithm. The whole framework was illustrated through simulation of BIP. To attest the stability of biped walking, interaction forces and friction cone have been researched by simulation studies.