基於多重感測之動力外骨骼機器人穿戴者意圖預測

Abstract

我們提出了一種用於週期性任務中輔助關節扭矩曲線的自適應學習的外骨骼意圖預測方法. 輔助裝置，例如外骨骼或假體，通常利用允許檢測或預測運動發作的生理數據。運動發作可以通過EMG或慣性傳感器在執行部位，骨骼肌上檢測到. 研究的最終目標是檢測和預測人類運動活動和姿勢，以一種可以改變外骨骼系統的運動活動的方式來發出輔助關節扭矩行為。進行實驗調查，將IMU傳感器放置在下肢位置，以獲得人體運動中的人體 - 機器人相互作用的方向，移動質心（COM）和速度變化，以了解人類運動中的肌肉活動. 力傳感器放置在腳底部，以獲得與外骨骼和人體運動對準的壓力中心（COP）和地面反作用力（GRF) 基於所獲取的數據，本地步態算法是為最大可能的行走模式而建立的. 我們提出的學習系統使用GAIT算法作為軌跡生成器，並使用線性回歸調製GAIT參數. 然後，將來，學習系統將與人機器人結構的動力學結合起來，將所期望的動力學改變為對主控制器的最終命令. 所提出的方法的優點是它不需要特定的生物力學模型，因為系統可以自適應地預測用戶在人和外骨骼機器人之間有效的人機器人交互的意圖.

We propose an exoskeleton’s intention prediction method for adaptive learning of assistive joint torque profiles in periodic tasks. Assistive devices, like exoskeletons or prosthesis, often make use of physiological data that allow the detection or prediction of movement onset. Movement onset can be detected at the executing site, the skeletal muscles, by means of EMG or inertial sensors. The ultimate goal of the research is to detect and predict the human movement activity and orientation signaling an assistive joint torque behavior in a way that the movement activity of the exoskeleton system can be modified. An experimental investigation is carried out with the placement of IMU sensors at the lower limb positions to acquire the orientation, shift in the center of mass (COM) and the change in velocity of the human-robot interaction to know the muscle activity during human locomotion. Force sensors are placed at the bottom of the foot to acquire the center of pressure (COP) and the ground reaction force (GRF) in alignment to exoskeleton and human locomotion. Based on the acquired data indigenous gait algorithm is built for the maximum possible walking patterns. Our proposed learning system uses GAIT algorithm as a trajectory generator, and parameters of GAIT are modulated using linear regression. Then, in the future, the learning system will be combined with the dynamics of the human-robot structure to alter the desired dynamics as the final command to the main controller. The advantage of the proposed method is that it does not require specific biomechanical models as the system can adapt itself to predict the intention of the user to have efficient human robot interaction between the human and exoskeleton robot.