Enable Fully Customized Assistance: A Novel IMU-Based Motor Intent Decoding Scheme

Chunzhi Yi, Shengping Zhang, Feng Jiang, Jie Liu, Zhen Ding, Chifu Yang, Huiyu Zhou

Abstract: Trustworthy human-exoskeleton interaction essentially relates to determining the assistive force profile. Current methods of decoding human motor intent enable the customized determination of the assistive force profile by providing limited information of human kinetics. In this article, we propose and validate a novel motor intent decoding scheme that can enable a fully customized assistive force profile, where only inertial measurement units (IMUs) are used. First, we improve the robustness of the IMU-based kinematic estimation by sampling IMU measurements that well meet the hinge-joint assumption, and by online calibrating axes’ direction in order to avoid the post-hoc analysis of joint axes’ directions during the determination of the body-fixed coordinate frame. Second, using the calculated kinematics as input, we develop a computationally efficient dynamic model, through which kinetics of users can be calculated in real time. Finally, we leverage a cable-driven ankle exoskeleton method to validate the assistive performance of our motor intent decoding scheme. We perform experiments on ten healthy subjects to evaluate the accuracy of our algorithm, and the change of metabolic rate and muscle efforts under the exoskeleton’s assistance. The results show the improvement from determining the assistive force profile by nominal curves and the feasibility of our algorithm.