Adaptive force control for walking robots with torque observer

Abstract

A novel control strategy for force control of walking robots is described. Force controls are important for walking robots because each leg must be allotted proper equal force (weight). The proposed control architecture consists of a position servo system with quick response, a torque observer, and a surface characteristics estimator. The control has two phases: a command is given to the position servo system to let the leg move to the specified position and the torque observer detects the contact force. A predictor algorithm estimates the relation between the position and the force by using a database constructed experimentally, and a new position command is generated on the basis of this relation. The total system acts like an adaptive controller, and each leg can support the specified weight on any surface.<>