Robotic Bin-Picking under Geometric End-Effector Constraints: Bin Placement and Grasp Selection

In this paper we demonstrate how path reachability can be taken into account when selecting among predetermined grasps in a bin-picking application, where grasps are supplied independently of the robot at hand. We do this by creating a map of the workspace to optimally place the bin with regards to the existence of an inverse kinematic solution and a collision-free path, a necessary condition for systems with obstructions in the workspace. Furthermore, we densely re-map this region and based on this map predict whether a grasp is reachable by the robot. Moreover, an algorithm is implemented to weight the grasps in terms of path existence, length and time consumption. The algorithm was tested with grasps generated by the neural network in simulation and the results indicate that faster picking can be achieved when taking path reachability into consideration.