Neural Network-Based Indirect Adaptive Force Control of a Two-Fingered Hand Exoskeleton Toward Robust Grasping Assistance

Grasping task is one of the crucial objectives in activities of daily living. However, elderly human subjects are facing significant challenges when attempting to perform grasping task. In this regard, a hand exoskeleton with a proper force control strategy is necessary to improve the performance of assistive technology. In this article, a lightweight, size-adjustable, underactuated, and force-controllable two-fingered exoskeleton prototype is developed for grasping assistance. A novel radial basis function network-based indirect adaptive force controller for robust grasping assistance is proposed along with the prototype design. We have conducted real-time grasping experiments on six elderly human subjects to verify the feasibility of the developed exoskeleton with a novel grasping force control strategy. Furthermore, we have performed force trajectory tracking experiment to validate the efficacy of the proposed force control scheme. Moreover, the robustness of the proposed grasping force control strategy has been validated through a disturbance rejection experiment. Extensive simulation and experimental studies with the developed kinematic model and feasibility tests involving elderly human subjects show that the newly developed hand exoskeleton with the proposed robust intelligent control strategy is efficient for object-grasping tasks aimed at the assistance of elderly human subjects.