Real-Time Control Strategy of Exoskeleton Locomotion Trajectory Based on Multi-modal Fusion

The exoskeleton robot is a typical man–machine integration system in the human loop. The ideal man–machine state is to achieve motion coordination, stable output, strong personalization, and reduce man–machine confrontation during motion. In order to achieve an ideal man–machine state, a Time-varying Adaptive Gait Trajectory Generator (TAGT) is designed to estimate the motion intention of the wearer and generate a personalized gait trajectory. TAGT can enhance the hybrid intelligent decision-making ability under human–machine collaboration, promote good motion coordination between the exoskeleton and the wearer, and reduce metabolic consumption. An important feature of this controller is that it utilizes a multi-layer control strategy to provide locomotion assistance to the wearer, while allowing the user to control the gait trajectory based on human–robot Interaction (HRI) force and locomotion information. In this article, a Temporal Convolutional Gait Prediction (TCGP) model is designed to learn the personalized gait trajectory of the wearer, and the control performance of the model is further improved by fusing the predefined gait trajectory method with an adaptive interactive force control model. A human-in-the-loop control strategy is formed with the feedback information to stabilize the motion trajectory of the output joints and update the system state in real time based on the feedback from the inertial and interactive force signal. The experimental study employs able-bodied subjects wearing the exoskeleton for motion trajectory control to evaluate the performance of the proposed TAGT model in online adjustments. Data from these evaluations demonstrate that the controller TAGT has good motor coordination and can satisfy the subject to control the motor within a certain range according to the walking habit, guaranteeing the stability of the closed-loop system.