Research on the human-following method, fall gesture recognition, and protection method for the walking-aid cane robot

Walking-cane robot is a hotspot in the field of human augmentation robots. Among them, the walking-aid cane robot has the way of use according to the patient’s habits because it does not interfere with the patient’s gait directly. The walking-aid cane robot is compact and flexible in movement, which is convenient for patients to walk and veer in daily life. This paper will design a walking-aid cane robot and propose methods of autonomous following and fall protection. The main work in this paper is as follows: 1. Robot’s control method of the autonomous following. The autonomous following of the robot is realized through the lidar. According to the point cloud data collected by the lidar, the human legs are separated from noise points and obstacles through a clustering algorithm, and the human point cloud is extracted through the HOG (Histogram of Oriented Gradients) and SVM (Support Vector Machine) methods. Then it uses PID (Proportion Integration Differentiation) control to the robot’s speed and smooth processing to the position. 2. Research on fall gesture recognition and protection methods. Through the OpenPose human gesture recognition repository, the key node information of the body in the camera image is analyzed in real-time to judge the falling trend of the body. The position that the robot needs to reach is calculated by the inclination of the body so that the robot can provide reverse support before the human body falls.