This thesis addresses two main issues of devices that assist human movements: One is stability of the coupled human-robot system, the other is how to make the human the master of the device. Firstly, passive control of exoskeletons with Series Elastic Actuators (SEAs) is investigated. SEAs decouple motor inertia from the human by a spring, reducing undesired interac-tion forces as a prerequisite for making the human the master. An important result is that if passivity is desired, the SEA cannot guide the limbs with a higher stiffness than that of the spring. Besides passivity analysis, which does not require a model of the human, also model-based stability analysis of human-robot systems is presented, explicitly addressing nonlinearity, time-variability, input saturation, and uncer-tainty. The example is a Hybrid Neuroprosthesis, which combines an exoskeleton with electrical stimulation of the human muscles. Robustness and performance of different controllers is compared. Conservative over- pproximation enables simple stability analysis of the complex structure, indicating that elastic coupling between human limbs and robot can cause instability. The required model of muscle recruitment dynamics is obtained via identification, using an inverted, anti-causal system description. Finally, a correlation- ased method is proposed to estimate desired motion for impaired or missing human limbs by complementing residual body motion. This way, the assistive device is reduced to a mere tool for the human, who regains control of lost motor functions. All methods are evaluated in simulations and in practical experiments with healthy subjects.

Категория: Научные статьи | Нет комментариев »

Комментарии