Heuristic-Based Ankle Exoskeleton Control for Co-Adaptive Assistance of Human Locomotion

Assisting human locomotion with exoskeletons is challenging, largely due to the complexity of the neuromusculoskeletal system, the time-varying dynamics that accompany motor learning, and the uniqueness of every individual’s response to device assistance. Assistance strategies designed to keep the human “in-the-loop” can help overcome many of these challenges. The purpose of this study was to develop a human-in-the-loop assistance strategy that uses co-adaptive control to slowly and continuously respond to biomechanical changes thought to encode the user’s needs. Online measurements of muscle activity and joint kinematics were used to guide the evolution of an exoskeleton torque pattern based on the following heuristics: 1) muscle activity that acts cooperatively with the exoskeleton indicates the user wants more torque; 2) muscle activity that acts antagonistically to the exoskeleton indicates the user wants less torque; and 3) torque should stop increasing if the user is not adapting. We applied our controller to tethered, bilateral ankle exoskeletons worn by naïve participants as they walked on a treadmill at 1.25 m $\cdot$ s−1 for 30 minutes. The evolved torque profiles reduced the root-mean-square of soleus muscle activity by 35±12% and metabolic rate by 22±8% compared to walking with the exoskeletons while they provided no torque. This was equivalent to a 9±12% reduction in metabolic rate when compared to normal walking. Furthermore, the algorithm was responsive to changes in each user’s coordination patterns. These results confirm the effectiveness of the controller and suggest a new approach to exoskeleton assistance aimed at fostering co-adaptation with the user. This technique might particularly benefit individuals with age-related muscle weakness.