Orthopaedic research studies the function, injury, and repair of the musculoskeletal system. Some orthopaedic research employs cadaveric testing;​ it allows many methods that are impossible in vivo, but it lacks the central nervous system to actuate a specimen. The joint motion simulator is an orthopaedic testbed developed to provide repeatable actuation of a cadaveric specimen. The control of joint motion simulators is complicated by nonlinearities, tendon actuation, and specimen variation.

This dissertation develops a control system for an elbow joint motion simulator. A combination of decoupling, mixed sensitivity robust control, and iterative learning control are used to track 100° sinusoidal flexion/extension and pronation/supination motions both individually and simultaneously. Small muscle co-contractions are maintained to prevent tendon slack. The control approach is applied to three systems of increasing complexity:​ flexion/extension controlled by the brachialis, flexion/extension controlled by the brachialis and triceps, and both flexion/extension and pronation/supination controlled by the biceps, brachialis, pronator, and triceps. All three levels of control are applied to a mechanical elbow and the two degree of freedom control is also applied to both a large and small cadaveric elbow. To compare with the state-of-the-art elbow joint motion simulator, joint position tracking error goals were set at 1.5° of root mean square error and 4° of maximum error;​ co-contraction moment error goals were set at 25 N-mm of root mean square error and 75 N-mm of maximum error for flexion/extension and 30 N-mm of maximum error for pronation/supination. All error goals were achieved for the first and second stages of testing. During the third stage of testing with the mechanical elbow and small cadaveric elbow, the tracking goals were achieved in all cases except during a flexion motion with fixed pronation angle. During the third stage of testing with the large cadaveric elbow, the tracking goals were achieved in all cases except the simultaneous pronation and extension motion and again during the flexion motion with fixed pronation angle.

This dissertation establishes a structured method of control synthesis for joint motion simulators devoid of the ad hoc methods often employed in joint motion simulator control.