Preventing Ankle Sprain: Integrating Preview Control Barrier Functions with Human Movement Primitive Prediction

• Published in: IFAC-PapersOnLine Vol. 58; no. 30; pp. 61 - 66
• Main Authors: Hwang, Seunghoon, Pati, Tarun, Lee, Hyunglae, Yong, Sze Zheng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2024
• Subjects: Assistive devices; Assistive robotics; Exoskeletons; Assistive robotics; Assistive devices; Exoskeletons
• ISSN: 2405-8963; 2405-8963
• DOI: 10.1016/j.ifacol.2025.01.157

This paper presents a safety controller for preventing ankle sprains, one of the most common musculoskeletal injuries, via a non-conservative control barrier function that incorporates uncertain preview/predictions of human ankle motion primitives. Specifically, our approach predicts the confidence interval of human ankle joint inversion/eversion angle trajectories during the stance phase using a probabilistic movement primitive (ProMP) and accounts for the prediction uncertainty when ensuring safety. Safety is guaranteed with high probability using an uncertain-preview control barrier function (uPrev-CBF) framework where the ProMP predictions, though uncertain, are leveraged as previewable disturbances to obtain a less conservative safety controller. Results from the ProMP model learning, based on data from four healthy subjects, demonstrate that the model effectively captures the variability in ankle inversion/eversion motion with high confidence. In addition, simulation results of the proposed safety controller based on uPrev-CBF show that ankle angle limit violations can be eliminated, effectively preventing ankle sprains.