Interactive Perception of Rigid and Non-Rigid Objects Regular Paper

This paper explores the concept of interactive perception, in which sensing guides manipulation, in the context of extracting and classifying unknown objects within a cluttered environment. In the proposed approach, a pile of objects lies on a flat background, and the goal of the robot is to isolate, interact with, and classify each object so that its properties can be obtained. The algorithm considers each object to be classified using color, shape, and flexibility. The approach works with a variety of objects relevant to service robot applications, including both rigid objects such as bottles, cans, and pliers as well as non-rigid objects such as soft toy animals, socks, and shoes. Experiments on a number of different piles of objects demonstrate the ability of efficiently isolating and classifying each item through interaction.