A modular and extensible framework for real and virtual bin-picking environments

Abstract

The number of industrial applications where a robot needs to unload disordered parts is increasing substantially. The usage of robot vision is highly preferred to obtain reliable results. In the past different algorithms for localizing objects as well as for motion planning to avoid collisions between gripper, object and environment were presented and even some commercial Bin-Picking systems are available. Nevertheless a realistic simulation taking account of various vision and robot systems and path planning strategies to predict cycle times is still not possible. In this paper a universal software framework with the focus on virtual Bin-Picking is presented. Utilizing a generic approach this framework enables the integration of various algorithms for object recognition, motion planning, different types of robots, grippers and vision systems. Thus an efficient simulation of different (virtual) Bin-Picking setups including equipment such as sensor devices or robot controllers within a virtual environment is possible. Furthermore the use of such a system allows the prediction of cycle times, percentage of tangible objects or testing of different Bin-Picking setups.