A Novel Gait Phase-based Control Strategy for a Portable Knee-ankle-foot Robot

This paper presents a novel control strategy of a portable knee-ankle-foot robot for overground gait training based on seven gait phases. Following assisted-as-needed (AAN) control strategies, the robot is able to provide assistance to the lower limbs in any gait phases, which deliver intensive training to certain lower-limb impairments. The gait phases can be detected in real time via inertia measurement unit (IMU) with hidden Markov model (HMM) and online Viterbi algorithm. The level of the robotic assistance adapts based on the kinematics of the lower limbs during overground walking. This control strategy has been tested on a healthy subject with an elastic bandage on the knee to simulate stiff knee condition of stroke patients. Adapting and increasing levels of assistance were provided to help the knee flex into a normal range. EMG profile of four major muscles on the assisted leg was collected. Experiment results indicate that the gait phase detection algorithm is accurate and robust and the robot is able to provide effective and synchronized assistance to the subject during overground walking.