Tactile Profile Classification Using a Multimodal MEMs-Based Sensing Module

Abstract

Robots are expected to perform complex dexterous operations in a variety of applications such as health and elder care, manufacturing, or high-risk environments. In this context, the most important task is to handle objects, the first step being the ability to recognize objects and their properties by touch. This paper concentrates on the issue of surface recognition by monitoring the interaction between a tactile probe in contact with a surface. A sliding motion is performed by a robot finger (i.e., kinematic chain composed of 3 motors) carrying the tactile probe on its end. The probe comprises a 9-DOF MEMs MARG (Magnetic, Angular Rate, and Gravity) sensor and a deep MEMs pressure (barometer) sensor, both embedded in a flexible compliant structure. The sensors are placed such that, when the tip is rubbed over a surface, the MARG unit vibrates and the deep pressure sensor captures the overall normal force exerted. The tactile probe collects data over seven synthetic shapes (profiles). The proposed method to distinguish them, in frequency and time domain, consists of applying multiscale principal components analysis prior to the classification with a multilayer neural network. The achieved classification accuracies of 85.1% to 98.9% for the various sensor types demonstrate the usefulness of traditional MEMs as tactile sensors embedded into flexible substrates.