UC Berkeley

Integrating an exoskeleton as the external apparatus for a brain-machine interface has the advantage of providing multiple contact points to determine body segment postures and allowing control to and feedback from each joint. Most current brain-machine interface studies use non-human primates, for example rhesus macaques (Macaca mulatta), as the research subjects. In order to develop an upper limb exoskeleton for macaques which can provide both data acquisition and motion actuation, this dissertation investigates the mechatronic considerations of developing such a device, including 1) the kinematic modeling and structural design, and 2) the actuator design and control.

An exoskeleton is a wearable robot, and is supposed to be attached to the user's body segments. Thus the kinematic structure needs to match the macaque's upper limb as closely as possible, which requires the exoskeleton to be compact and singularity free in workspace, as well as have biomorphic nature joints and firm attachments. In order to provide sufficient output torque and guarantee the user's safety, the actuators of the exoskeleton should be backdrivable, of high power-to-weight ratio, and capable of executing compliant actuation algorithms.

Based on the above design requirements, this dissertation presents the development process of an upper limb exoskeleton for macaques from kinematic modeling and analysis, passive exoskeleton prototyping and animal training, compliant actuator design and control, to completion of an actuated exoskeleton and system integration with a brain-machine interface. First an upper limb exoskeleton model is proposed with a redundant shoulder joint to achieve improved manipulability than conventional designs. The advantageous features of the proposed exoskeleton model are demonstrated by a series of kinematic analysis. Then a passive upper limb exoskeleton is fabricated for kinematic model validation, motion characterization and animal training purposes, and its effectiveness is demonstrated by the animal tests. In order to obtain a compact, powerful, backdrivable and torque-reflecting actuator for the actuated exoskeleton, a cable-driven series elastic actuator is designed, and an interactive impedance control algorithm is proposed and experimentally validated. On top of the development of the passive exoskeleton and the impedance-controlled actuator, an actuated multi-degree of freedom upper limb exoskeleton is developed and integrated with a brain-machine interface, and the effectiveness of the proposed exoskeleton system is finally supported by pilot animal tests.