Vision Sensor Based Safety Monitoring In Human And Robot Coexistence Environment

Robot will be a part of our society in the future and they are expected to interact and collaborate with human in daily life and production,where requires them enable to adapt the dynamic scenarios in real time.The most significant point is how to guarantee the human and robot can coexist with safe.Because of it is an uncertain environment where introduced by the variability inherent in human motion and preferences,therefore,some additional sensors needed to detect the human or some unknown objects which may conflict with the robot motion and also some real time collision avoidance algorithms should be developed to control the robot away from the object.To approach the issue of human and robot coexistence,a system with depth camera to monitor the surrounding of the robot,which can capture the dynamic objects including human or some other objects,in addition,an algorithm of avoiding obstacle in real time is proposed,thus the robot is capable of keeping a certain safe separation with the obstacle.All in all,our work can be divided into three parts as follows.Firstly,utilizing depth camera to detect the object with potential risk of colliding with the robot,but before that we use OpenGL to construct a scene where only the monitored robot is defined to filter the robot in depth image to allow stable human detection or motion capture.Afterwards,the static background can be removed by subtracting a depth image stored before without any humans in current view,and then translating the depth image into point clouds to find out the closet cluster to the robot.Secondly,based on the principle of dynamical system modulation we developed a real time safe motion generation algorithm utilizing the repulsive vector generated between control point and the point clouds of obstacles to construct the modulation matrix that can be modified the system state of the robot to away from the obstacle.Thirdly,considering the robot body collision avoidance by null space control,which is incorporates the method of the repulsive vector.To address the issue of balancing the end-effector task and the robot body collision avoidance task we proposed a modified task weight function which takes those two tasks constrains into consideration.Experiments with a UR5 robot interacting with a human under the surveillance of two Xtion depth sensors are conducted to showcase the effectiveness of our proposed method.Moreover,a redundant robot,KUKA iiwa,is also monitored by a depth camera to verify the algorithm of task transition can make the robot link avoid obstacles,and our system can guarantee the human and robot coexistence.