Model-based Controlling of Robots for In-plant Transportation of Cylinder-shaped Objects by Throwing and Capturing☆

Abstract Factories are confronted with an immense stress of competition. Hence, their performance parameters, such as throughput rates or processing times, have to be continuously improved. Processes have to be designed providing a maximum of efficiency without suffering loss of flexibility especially in consideration of shortened product lifecycles and increasing product variants. In this context, also internal logistic processes, such as in-plant transportation of material, have to be considered by factory planners. In our research study, a new approach is investigated, which considers transportation of objects within production systems by means of automated throwing and capturing. This visionary approach basically aims at shortening times of transport, reduction of expensive conveying systems and improvement of flexibility in material flow. In the context of this research, a robot system has been implemented, which is capable of automatically throwing and capturing stably flying cylinder-shaped objects over a distance of three meters. Cylinder-shaped objects with varying masses and geometric dimensions have been chosen for investigation since these are widespread in mechanical engineering (e.g. axles, pins or bolts). Cylinders are accelerated in their axial direction by a throwing robot with a linear axis. A second robot captures them by means of a gripper. In order to have this done smoothly th e movement of the gripper must be aligned to the position, orientation and velocity of the flying cylinder at the capturing point. For appropriate control of the robots, a model is introduced, which considers the aerodynamics of the thrown cylinder, in order to compute its pose and velocity along the trajectory versus time. Experimental results demonstrate that the position and velocity can be calculated with good accuracy. Due to imperfections, such as vibrations of the throwing robot, the orientation of the cyli nder, however, can considerably deviate from the calculated value, which may lead to failure of capturing. Thus, the movement of the capturing robot has to be adapted in real-time in order to compensate these deviations. To detect these deviations, the system must be equipped with a sensor system. With this sensor system, the pose of the flying cylinder is measured approximately one meter before capturing. An algorithm has been developed, which enables an online prediction of its pose at capturing point based on the measured data thus allowing the robot to be guided to this pose timely.