Monitorování sdíleného pracovního prostoru mezi člověkem a cobotem

Abstract

The dissertation deals with the monitoring of the workspace of a robotic workstation designed for collaboration between a human and a collaborative robot (cobot). There is a large increase in the deployment of collaborative robots within Industry 4.0. Even though these robots are adapted to collaborate and share workspace with humans, the efficiency of the workplace is not as high as that of standard industrial robots. Since the cobot performs its predefined task without regard to the worker and objects in the workspace, collisions can occur. Due to the characteristics of cobots, then, such a collision is normally safe and will not injure the worker or damage the equipment. However, the consequence of even a minor collision is extended work cycle of the workplace caused by an emergency stop of the cobot. The safety features available today do not allow sensing of the entire dynamically changing environment of robotic workplaces to such an extent that the robot trajectory can be re-planned in real time, and thus avoid collisions between the worker or objects. The introductory part of this thesis deals with the analysis of the current state of the art in 3D vision, calibration, environmental filtering, and optimization of camera placement in the workplace. Based on the stated objectives of the dissertation, the possibilities of improving the accuracy of estimating the position of cameras in space are then investigated. A newly designed 3D gridboard for camera positioning has been implemented on industrial and service applications. Furthermore, in order to use the camera data efficiently for the manipulator control system, a filtering technology is developed to remove unnecessary data in real time and to transmit only the necessary information to the higher-level control system. By using a distributed structure, real-time sensing and filtering of the monitored area was achieved, regardless of the number of cameras in the system, without slowing down the refresh rate of the scene. Furthermore, a methodology for deploying cameras in a workplace was established to effectively capture the hazardous area where a cobot collision is imminent. Finally, this issue is experimentally verified and implemented in a real control system for dynamic obstacle avoidance.