Humanoid Robot Locomotion Control based on Perceptive Model

Abstract

A humanoid locomotion control method with human-like perceptive model is described in this paper. Traditionally the locomotion problem of biped robot is divided into temporal periodic gait generation and real-time stabilizing control. However due to unmeasurable accumulated errors from simplified dynamic model and other uncertain disturbances, the spatial and temporal deviation from the motion plans askes for a new perceptive locomotion planning and control system which is decoupled with time, rather than adhering to a rigid time schedule regardless of current dynamic balance state. This method also gives the right of interpretation of non-time motion plan of gait to the motion control layer based on a designed perceptive reference, which is directly ensure dynamic balancing and high response rate to the errors. The simulation and experiment on a full-size humanoid robot show the success of the method.