Promoting Co-Adaptation in Human Interaction with Powered Upper Limb Exoskeletons

Abstract

Powered robotic exoskeletons offer great promise in assisting elderly or disabled individuals with daily activities or augmenting healthy individuals in performing labor-intensive tasks. Yet, one primary challenge for these powered devices is to achieve seamless cooperation between the human user and the exoskeleton. Fluent human-exoskeleton cooperation mandates mutual adaptation. Humans need to adapt to it as the support from the exoskeleton alters movement dynamics. Similarly, the exoskeleton must also adapt to the evolving behaviors of its users. However, current approaches often lack adequate guidance to facilitate human learning in exoskeleton usage, and many exoskeleton controllers overlook the user’s evolving behaviors during the learning process. Moreover, the dynamics of co-adaptation between the human and exoskeleton are not thoroughly understood. The dissertation delves into each of these challenges through three primary projects.

In the first project, I examined the impact of EMG biofeedback on the use of EMG-based powered upper limb exoskeletons in a tracking task. We found that EMG biofeedback did not lead to large differences in muscle effort reduction (overall or during hold periods) or task accuracy, possibly due to the need for individuals to have additional support to acquire an appropriate exoskeleton motor program and learn how to effectively utilize the biofeedback information. Nevertheless, biofeedback may improve participant satisfaction with exoskeleton usage, which is a crucial factor for encouraging long-term use. In the second project, I introduced an adaptive controller capable of monitoring user adaptation and utilizing this information for intention classification. The adaptive controller yielded notable enhancements in intention classification accuracy and reduced muscle effort during movement initiation in a target position matching task. In the third project, I integrated EMG biofeedback with the adaptive controller to promote co-adaptation within the humanexoskeleton team. I also investigated the impact of biofeedback training on novice users’ ability to utilize biofeedback information effectively. The results demonstrated that training enabled users to benefit from biofeedback, and the co-adaptation process led to improved exoskeleton usage.

Taken together, these projects provide critical insights into promoting co-adaptation in humanexoskeleton interaction. They offer an understanding of human and exoskeleton adaptation independently while also unveiling the interactive dynamics of co-adaptation, advancing the field toward more effective and intuitive exoskeleton systems.