Lower-Limb Movements Prediction Based on IMUs for Exoskeleton Via Adaptive Temporal Movement Primitives

A novel approach to predict human lower limb movement (LLM) based on adaptive temporal movement primitives (ATMPs) which are inspired by the principle of alpha motor neurons is proposed in this study. The adaptive algorithm is presented that can adjust to different locomotion modes dynamically. The LLM prediction method is verified on an experimental load-carrying lower extremity exoskeleton and demonstrates its accuracy under complex locomotion modes. The performance of the exoskeleton equipped with the method is evaluated in a field environment with seven typical and three transitional locomotion modes. The proposed method achieves a prediction horizon of 148 ms and a prediction root-mean-square error of 4.25%. The experiments confirm the adaptability and accuracy of the ATMPs-based method for LLM prediction.