パワーアシストロボットのAssist-as-Needed適応速度場制御

抄録

This paper proposes an AAN VFC (Assist-as-Needed Velocity Field Control) approach for rehabilitation robots. Proposed controller includes an adaptive NN (neural network) term to compensate for the unknown dynamics of the system and the weight matrix update law of the adaptation NN term involves a forgetting factor which reduced the control effort for small tracking error. The dead-zone property together with the forgetting factor of the NN gives a liberty (a free torque tunnel) to the user to move the target limb with a velocity inside a velocity interval around the desired velocity. This property leads to the AAN property. The controller does not require the dynamic model of the system, and also delivers a priori bounded command. The latter has been featured by means of utilizing a saturated function in the proportional-type feedback term and a projection operator in the NN update law. The stability of the closed-loop system is studied well, and the performance of the controller is evaluated through experiments conducted on a lower-limb exoskeleton (TTI-Knuckle 1).