A Binocular Vision Based Intelligent Upper Limb Exoskeleton for Grasp Assisting.

Assisting patients with upper limb motor dysfunction to complete activities of daily living can improve their quality of life. This paper develops a binocular vision-based upper limb exoskeleton to assist grasping. The mechanism adopts a modular design to reduce the complexity of the joint. The cable-driven actuators were used to improve system compliance. A binocular vision-based detection and location system was developed, which adopted a depth camera and deep learning detection algorithm to identify and locate objects. The frame of the target object detected by the master computer was sent to the trajectory controller, then the trajectory controller controlled the cable-driven actuator to approach the target object by calculating the inverse kinematics of the multi-joint mechanism. Grasping experiments demonstrated that the system could effectively assist patients with upper limb dysfunction in recognizing and locating target objects.