Full-body motion control integrated with Force Error Detection for wheelchair support

Abstract

We propose a full-body controller system for a humanoid robot to achieve wheelchair support. There are two main problems in pushing control. One problem is that the controller system should decide pushing strategies based on whether the humanoid robot should adapt to the reaction forces from the pushed wheelchair or resist against them. Another problem is that the arms's reachabilities decrease while walking because of upper body motion. We resolve these two problems by extending the full-body controller by detecting pushing error and by determining the place where the humanoid robot stands based on off-line computation of the arms' reachabilities. Our system includes estimation of the friction forces based on on-line reaction force measurements. If the force error becomes large, the system suspends pushing control to switch a pushing strategy. We also show experimental results on the HRP-2 humanoid robot. In the experiment, the humanoid robot navigated a heavy wheelchair in environment including several furniture.

Can I lift it? Humanoid robot reasoning about the feasibility of lifting a heavy box with unknown physical properties

“…There exist many studies focusing on humanoid robots moving large and cumbersome objects, such as pushing heavy objects [11], moving wheel chairs [12], moving heavy boxes via a series of pivoting motions [13], carrying a heavy object while keeping balance [14], transporting heavy objects in cluttered space [15], etc. The problem of lifting heavy objects is also mentioned in the literature.…”

Control strategy and implementation for a humanoid robot pushing a heavy load on a rolling cart

“…The authors assessed the importance of integrating the force generated by the pushed object in the walk controller to ensure stability. The proposed system is based on an online friction estimator [5] that is able to estimate the friction forces generated by the object using measurements from force/torque sensor located in the wrists and ankles of the robot. Then, the desired ZMP trajectory is modified and supplied to the preview controller to consider the external force.…”