Online Gait Trajectory Planning of a Knee Exoskeleton Robot Based on Kernelized Movement Primitives

In order to improve the adaptability of an exoskeleton robot and the safety of human-exoskeleton interaction, an online gait trajectory planning method based on kernelized movement primitives (KMPs) is proposed, which can adjust the gait trajectory in real time according to the user's gait state. The algorithm is demonstrated on a knee joint exoskeleton. The result shows that the proposed algorithm can adjust the shape parameters online according to a few training samples, to ensure that the predicted trajectory is similar to the original trajectory. It can also generate a trajectory according to user's state change by detecting gait phase in real time and adjusting the gait trajectory online at the desired point.