Sensor-based Real-timeControl of Industrial Robots

Abstract

This PhD thesis presents topics related to sensor-based real-time robot control in applied automation.\n\nWhile in the past industrial challenges often could be solved with traditional robot programming methods, the industry looks more and more towards the automation of complex tasks that need real-time sensor-feedback in the control loop of the robot. This is especially important in applications where the workpiece shape or characteristics are uncertain or unknown. When working on such workpieces, the movement has to be adjusted in real-time, ie. while the robot is moving. To achieve this, sensors are added in the control loop of the robot.\n\nOne challenge for the implementation of industrial solutions including real-time control is the lack of appropriate interfaces in most commercially available robots. When such interfaces are not present, in-house developed solutions can be used to access the low-level system of the robots, at least for research applications. These usually require modifications of the hardware, the software platform or both. Often, interfaces are implemented that allows for parts of the control loop being moved to an external PC. In this thesis, real-time interfaces for two different industrial manipulators are presented that are used for lab experiments and demonstrators. The real-time interfaces work position-based and have update frequencies of about 100 Hz. Ethernet UDP is used to communicate with the external controller platform. Experiments are presented that measure the delays in the low-level systems of the robots.\n\nTo externally control the movement of the real-time controlled robots from an application controller, a trajectory generator is needed. This thesis gives an overview of three different on-line trajectory generators that were in-house developed. The real-time capabilities of the presented trajectory generators were analyzed in order to ensure that the requirement for response times of the real-time interfaces are met.\n\nBased on a presented real-time interface and trajectory generator, a test platform was built. The test platform demonstrates tracking of a line that is sketched on the workpiece. The robot tool is controlled to keep a fixed distance between tool and workpiece, while maintaining an orientation perpendicular to the workpiece surface. The line tracking is done using a line-of-sight based control method. Experiments are presented, measuring the delays in the robot-sensor system.\n\nThe main part of this thesis is a presentation of an automated robotic sewing cell that demonstrates a case of sewing for the furniture industry. The system is able to sew together two parts with slightly different shapes. A two-robot solution is presented that controls the work pieces independently during the sewing operation. A force sensor is integrated in the control system to keep a constant sewing force. The seam allowance is controlled by an edge control system based on optical sensors that are mounted on the sewing machine. The real-time capabilities of the system are analyzed. Experiments are presented showing the feasibility of the presented control methods. The seam quality was evaluated by manual inspection and was found to be adequate. Further steps were identified that are necessary to include the demonstrator in an industrial setup, mainly corner matching, sewing of the last few centimeters and the material handling before and after the sewing operation.