Safe height estimation of deformable objects for picking robots by detecting multiple potential contact points

Abstract

Object sorting in logistics warehouses is still carried out manually, and there is a great need for automation with arm robots. It is desirable that target objects be carefully placed in situations where careful handling of products is important. We propose a method for estimating the height of picked object with a single depth camera to achieve precise placing of items such as stacking, especially for objects that are deformable, e.g., bags. The proposed method detects multiple potential contact points of a picked object to estimate the appropriate height to place the object using the point-cloud difference before and after picking. The validity of the proposed method was verified using 26 cases in which deformable objects were placed inside a container, and it was confirmed that object-height estimation is possible with an average error of 3.2 mm.