Pose-Based Tactile Servoing: Controlled Soft Touch using Deep Learning

TL;DR

This paper introduces pose-based tactile servo control using deep learning for soft tactile sensors, enabling robots to perform precise and robust manipulation of complex objects through controlled soft touch.

Contribution

It presents a novel tactile servoing method embedding a pose estimation model within a control loop, demonstrated with the BRL TacTip sensor and CNNs insensitive to shear.

Findings

Achieved robust motion control over complex 3D objects.

Demonstrated accuracy and insensitivity to shear in tactile perception.

Linked tactile servoing concepts to visual servoing for improved dexterity.

Abstract

This article describes a new way of controlling robots using soft tactile sensors: pose-based tactile servo (PBTS) control. The basic idea is to embed a tactile perception model for estimating the sensor pose within a servo control loop that is applied to local object features such as edges and surfaces. PBTS control is implemented with a soft curved optical tactile sensor (the BRL TacTip) using a convolutional neural network trained to be insensitive to shear. In consequence, robust and accurate controlled motion over various complex 3D objects is attained. First, we review tactile servoing and its relation to visual servoing, before formalising PBTS control. Then, we assess tactile servoing over a range of regular and irregular objects. Finally, we reflect on the relation to visual servo control and discuss how controlled soft touch gives a route towards human-like dexterity in robots.