Distributed Control and Safety System for a Rehabilitation Arm Exoskeleton

Abstract

A modular control and safety architecture is described for a robotic arm exoskeleton designed for shoulder rehabilitation. The exoskeleton joints are partitioned into anthropomorphic sets based on the desired exercise protocol, which are then commanded using separate control modules. A composite control software architecture is used to enable simultaneous operation of joint sets in either impedance or admittance control modes. Preliminary hazard and fault tree analyses identified three major hazards, and a combination of redundant sensing, hardware limits and software safety checks were used to produce a single fail-safe design. A quantitative fault tree analysis was also performed to assess the risk of a hazard due to a failure. The application of this approach to the MGA Exoskeleton was shown to improve the safety of the overall system.