Environmental perception and joint trajectory generation for transfemoral prosthesis

Abstract

Recent wearable robot technology has made significant advancements aimed at enhancing the mobility of lower-limb amputees and ndividuals with walking disabilities. This technology is designed to facilitate smooth movement not only on level ground but also on stairs and ramp. However, previous studies have primarily focused on steady state of gait in environmental conditions, limiting the robot’s ability to provide appropriate assistance during transitions between different terrains. In this study, we integrate depth cameras and IMU sensors to perceive the environment using visual inertial odometry and elevation mapping, and a stride model to detect the user’s walking intentions during terrain changes. Additionally, we utilize the ProMPs (Probabilistic Movement Primitives) model to predict the walking trajectory during the swing phase considering the height of the stairs and the state of the user when transitioning from level ground to stairs. Experimental results demonstrated that the proposed environmental perception model achieved over 95 % accuracy in outdoor environments. In the laboratory, the model accurately recognized the user’s intentions with more than 94 % accuracy for obstacles including curbs and stairs of various heights. For other environments, it recognized walking intentions with an accuracy of over 80 %. Moreover, the validation of the joint trajectory model through simulations showed success rates of 71.88 % and 97.92 % using actual and estimated thigh angles, respectively. This research validates the potential of a model that accurately predicts walking intentions and provides appropriate trajectories for transfemoral prosthesis used in various terrains.