View Planning for Randomized Bin-picking

Abstract

This paper proposes a method for view planning to be used for randomized bin-picking. Assuming a 3D sensor attached at the wrist, we consider obtaining the pose of 3D sensor minimizing the occuluded area of the view captured during the previous picking trial. Then, by comparing the current and the previous view, we consider identifying the pose of object only where the difference between two view is large. The effectiveness of the proposed method is verified by experiment of using a dual-arm manipulator.