Robust Markovian Impedance Control applied to a Modular Knee-Exoskeleton

Abstract

Abstract Lower limb exoskeletons have improved mobility and safety during gait rehabilitation. Joint actuators can be programmed to produce sufficient joint torque to promote human movement. However, the mechanical impedance of the human joints changes constantly to maintain a stable interaction with the environment during walking. These continuous changes introduce nonlinearities and uncertainties that alter abruptly the dynamics of the human-robot interaction, which can destabilize the control system. In this paper, an impedance control approach under explicit Markovian torque control architecture is developed, considering the variable human impedance parameters as parametric uncertainties. As the time-varying human dynamics during walking depends on the quasi-cyclic gait phase transitions, we defined five Markovian operation modes to describe the human-robot interaction during walking. Additionally, impedance parameters of the human knee joint were estimated using an ensemble-based method. Experimental results of the proposed control scheme on a knee-exoskeleton driven by a series elastic actuator show that our proposal guarantees stability and high performance despite the stochastic uncertain human impedance behavior throughout the gait cycle.