Laboratory evaluation of a unified theory for simultaneous multiple axis artificial arm control

This paper reports on the application of a "postulate-based" control method for multi-axis artificial arm control. This method uses shoulder muscle EMG's as control sites, but, unlike previous techniques, the theory is the first that can be rigorously defined in terms of musculoskeletal anatomy, EMG muscle-force relationships, EMG transmission characteristics, muscle recruitment, limb dynamics and normal motion constraints. The control theory results in a deterministic, mathematically expressible set of controller equations, which use the vector of natural limb torques estimated by shoulder EMG signals and a "constraint" for input. The output of the controller equations is a vector of prosthetic torques to be applied to the artificial limb. We report on the implementation of the theory up to the point of laboratory feasibility trials of actual simultaneous above-elbow amputee control of elbow flexion and humeral rotation. Implementation of the theory required: 1) deviation of the controller equations from Newton's dynamic equations of motion into controller form in conformity with the postulate theory; 2) development of a methodology for estimating natural musculoskeletal torques from EMG signals; 3) hardware and software for experimental testing with actual closed loop amputee control of the prosthesis; and 4) a methodology for evaluating the performance of the prosthesis relative to both alternative prosthetic systems and the natural arm. These tasks were completed and simultaneous multiple-axis control of a prosthetic arm was accomplished by both amputee and nonamputee subjects. Key questions of control compatibility, naturalness, stability, and performance evaluation relative to other prostheses and the natural arm were addressed. Various problems are discussed with the conclusion that this method, despite some difficulties, holds great promise as a practical rehabilitation tool.