Design optimization and control of a parallel lower-arm exoskeleton

Official URL: http://192.168.1.20/record=b1228175 (Table of Contents)

Wearable force feedback robotic devices, haptic exoskeletons, are becoming increasingly common as they find widespread use in medical and virtual reality (VR) applications. Allowing users to mechanically interact with computationally mediated environments, haptic exoskeletons provide users with better “immersion” to VR environments. Design of haptic exoskeletons is a challenging task, since in addition to being ergonomic and light weight, such devices are also required to satisfy the demands of any ideal force-feedback device: ability withstand human applied forces with very high stiffness and capacity to display a full range of impedances down to the minimum value human can perceive. If not properly designed by taking these conflicting requirements into account, the interface can significantly deteriorate the transparency of displayed forces; therefore, the choice of the kinematic structure and determination of the dimensions of this kinematic structure have significant impacts on the overall performance of any haptic display independent of the control algorithm employed. In this thesis, we first propose a general framework for optimal dimensional synthesis of haptic interfaces, in particular for haptic interfaces with closed kinematic chains, with respect to multiple design objectives. We identify and categorize the relevant performance criteria for the force feedback exoskeletons and address the trade-offs between them, by applying a Pareto-front based multi-objective design optimization procedure. Utilizing a fast converging gradient-based method, the proposed framework is computational efficient. Moreover, the approach is applicable to any set of performance indices and extendable to include any number of design criteria. Subsequently, we extend this framework to assist the selection of the most appropriate kinematic structure among multiple mechanisms. Specifically, we perform a rigorous comparison between two spherical parallel mechanisms (SPMs) that satisfy the ergonomic necessities of a human forearm and wrist and select the kinematic structure that results in superior performance for force-feedback applications. Utilizing the Pareto optimal set of solutions, we also assign dimensions to this mechanism to ensure an optimal trade-off between global kinematic and dynamic performance. Following the design optimization phase, we perform kinematic and dynamic analyses of the SPM-based exoskeleton in independent coordinates to facilitate efficient simulation and real-time implementation of model based controllers. We decide on the hardware components considering human wrist torque and force limits, safety and ergonomy constraints, and present the CAD model of a prototype of the exoskeleton. Finally, we implement model based task-space position and impedance controllers in simulation and present the results of them.