Enhancing Exoskeleton Assistance Flow Control by Leveraging Human Variability and Probabilistic Movement Primitives

Exoskeletons have the potential to significantly reduce users' physical effort and decrease the risk of work-related musculoskeletal disorders by providing robotic assistance. Human motor control is governed by a complex interaction between the central nervous system, the musculoskeletal system and the environment, generally leading to relatively fast, precise, smooth, and efficient movements. As a result, designing exoskeletons that provide assistance without being perceived as disruptive is challenging; they must be capable of predicting and adapting smoothly to the user's intended movements. In this paper, we propose leveraging the inherent variability in human movement to enhance the acceptability and comfort of upper-limb exoskeleton assistance. For that purpose, we use demonstrations from the human user and an online prediction method to infer their trajectory continuously, and we use an adaptive flow controller to adjust the assistance accordingly. We then increase the level of guidance and reduce correctivity of the exoskeleton when the users normally exhibit high variability, taking advantage of the higher tolerance to deviations from the planned trajectory in these phases to provide a stronger guidance while accommodating these deviations with higher compliance. Our approach was tested with the ABLE7D upper-limb exoskeleton during reaching tasks. The results demonstrate that our method can effectively reduce the user's physical effort, notably by lowering muscle co-contractions, compared to a classical data-based solution, while providing a comfortable level of assistance.