Step modeling and safe path planning for a lower limb exoskeleton

Abstract

The walking experience in environments with obstacles is a challenge for patients with lower limb pathology. A transparent exoskeleton is an interesting solution since it guarantees the performance of autonomous motion. In this paper, we present a new method to detect and model steps using point cloud data to find a feasible path for the exoskeleton to perform. We use a RGB-D sensor to obtain depth information and perform a scene segmentation strategy. Next, we classify the different detected elements either as a floor, step or obstacle and then use a path planning method to find a collision-free path. Experiments show that the system accomplished satisfactory results for the presented scenarios.