Ensuring safety in human-robot coexisting environment based on two-level protection

Purpose - The purpose of this paper is to provide a novel methodology based on two-level protection for ensuring safety of the moving human who enters the robot's workspace, which is significant for dealing with the problem of human security in a human-robot coexisting environment. Design/methodology/approach - In this system, anyone who enters the robot's working space is detected by using the Kinect and their skeletons are calculated by the interval Kalman filter in real time. The first-level protection is mainly based on the prediction of the human motion, which used Gaussian mixture model and Gaussian Mixture Regression. However, even in cases where the prediction of human motion is incorrect, the system can still safeguard the human by enlarging the initial bounding volume of the human as the second-level early warning areas. Finally, an artificial potential field with some additional avoidance strategies is used to plan a path for a robot manipulator. Findings - Experimental studies on the GOOGOL GRB3016 robot show that the robot manipulator can accomplish the predetermined tasks by circumventing the human, and the human does not feel dangerous. Originality/value - This study presented a new framework for ensuring human security in a human-robot coexisting environment, and thus can improve the reliability of human-robot cooperation.