| Resistance exercise has been reported with positive impacts on rehabilitating patients after traumatic injuries or with neurological conditions. Since the muscle force is not constant, a dynamic relationship does exist among muscle force, joint position, and muscle contraction speed. The variable resistance exercise accounting for such relationship should achieve optimal muscle strengthening effects. To enable functional resistance exercise in the patient's own home, we have developed a group of variable resistance exercise devices (VRED) in the form of knee braces and exercise chairs for muscle strengthening.; In this thesis, we discussed the hardware and software design of VRED, paid attention to a portable, intelligent, and functional exercise device with friendly human-machine interface. VRED provides a versatile platform that can be used to investigate the optimal exercise protocol both in the clinical and home-based setup. The software of VRED is programmed with LabVIEW. VRED is capable of isometric setup, optimal resistance profile exercise, machine adjust motion, and human control motion exercise. The resistance force of VRED comes from damper with magnetorheological fluids, that change flow viscosity in response to the applied magnetic field in milliseconds. We studied the optimal design of MR damper that minimizes the damper weight as well as satisfies the exercise force requirement.; Next we investigated the supervisory control of resistance force based on a nonlinear model with parametric uncertainty. Supervisory control is studied because of its simple formulation to tackle the model uncertainty. It also adapts to the portable application of VRED with limited computation power. We derived the control algorithm, analyzed the closed-loop stability, verified the control in both simulation and experiments. Then, adaptive control is studied to regulate the exercise motion considering the human joint dynamics and muscle force dynamics. Continuous-time adaptive control algorithms are formulated separately for excitation coil current loop, resistance force loop, and motion control loop. Simulations are performed to verify the effectiveness of algorithms. Experiments show the continuous-time adaptive control design is good for low-speed and less variable resistance exercise. Finally we also investigated the discrete-time adaptive control and system identification method in resistance force tracking to improve the performance of resistance exercise with limited sampling frequency. We found the algorithm is feasible and maintains the satisfactory performance under various exercise circumstances in both simulations and experiments. |
