Object mapping, recognition, and localization from tactile geometry

Abstract

Abstract-We present a method for performing object recognition using multiple images acquired from a tactile sensor. The method relies on using the tactile sensor as an imaging device, and builds an object representation based on mosaics of tactile measurements. We then describe an algorithm that is able to recognize an object using a small number of tactile sensor readings. Our approach makes extensive use of sequential state estimation techniques from the mobile robotics literature, whereby we view the object recognition problem as one of estimating a consistent location within a set of object maps. We examine and test approaches based on both traditional particle filtering and histogram filtering. We demonstrate both the mapping and recognition / localization techniques on a set of raised letter shapes using real tactile sensor data.

Iterative Closest Labeled Point for tactile object shape recognition

“…In a similar manner, relying on the locations of the contact points and hand pose configurations, a polyhedral model was derived to recover the shape of an unknown object in [9]. More recently, Pezzementi et al proposed a method to mosaic tactile measurements to construct an object representation [10]. Meier et al [7] applied Kalman filters to generate 3D representations of unknown objects from contact point clouds collected by tactile sensors and classified the objects with the ICP algorithm.…”

Localization and manipulation of small parts using GelSight tactile sensing

“…Platt et al use tactile maps to localize distinctive haptic features in soft materials [6]. Pezzementi et al use tactile models in order to classify and localize objects (for example, raised letters from a children's play set) [15]. Another important area of related work has to do with other tactile sensors that measure deformation in a deformable membrane.…”

Vibrotactile feedback for conveying object shape information as perceived by artificial sensing of robotic arm

“…However, most of their work is based on simulation results and has not been applied to practical cases for field testing. Pezzementi et al (2011) has considered object shape reconstruction and recognition as a search problem in continuous space. Simulation based experiment is done to identify complex shapes using bagof-features framework, following which the framework is tested to identify raised alphabets thereby validating their proposed work but only in a limited environment in practical scenario.…”

“…Ibrayev and Jia (2004) have derived differential invariants of low-degree polynomial curves which are considered to localize the object shape relative to the hand. Pezzementi et al (2011) has formed mosaic of tactile images and used state estimation techniques for mapping, identifying and localizing raised alphabets. Pezzementi et al (2011) has demonstrated haptic exploration of an object as a search problem in continuous space to explore and construct its appearance.…”

“…Pezzementi et al (2011) has formed mosaic of tactile images and used state estimation techniques for mapping, identifying and localizing raised alphabets. Pezzementi et al (2011) has demonstrated haptic exploration of an object as a search problem in continuous space to explore and construct its appearance. This shows that research on object shape recognition has been in focus in various fields since its inception till date.…”