FeelAnyForce: Estimating Contact Force Feedback from Tactile Sensation for Vision-Based Tactile Sensors

TL;DR

This paper introduces a vision-based tactile sensing method that estimates 3D contact forces over a wide range with high accuracy, generalizing across different sensors and objects, enabling applications like weighing and delicate object manipulation.

Contribution

It presents a novel transformer-based force regression model trained on extensive data, with a calibration procedure for sensor adaptation and demonstrated effectiveness on real-world tasks.

Findings

Achieves 4% mean absolute error on unseen objects

Generalizes across different tactile sensors

Enables real-world force-based manipulation tasks

Abstract

In this paper, we tackle the problem of estimating 3D contact forces using vision-based tactile sensors. In particular, our goal is to estimate contact forces over a large range (up to 15 N) on any objects while generalizing across different vision-based tactile sensors. Thus, we collected a dataset of over 200K indentations using a robotic arm that pressed various indenters onto a GelSight Mini sensor mounted on a force sensor and then used the data to train a multi-head transformer for force regression. Strong generalization is achieved via accurate data collection and multi-objective optimization that leverages depth contact images. Despite being trained only on primitive shapes and textures, the regressor achieves a mean absolute error of 4\% on a dataset of unseen real-world objects. We further evaluate our approach's generalization capability to other GelSight mini and DIGIT sensors, and propose a reproducible calibration procedure for adapting the pre-trained model to other vision-based sensors. Furthermore, the method was evaluated on real-world tasks, including weighing objects and controlling the deformation of delicate objects, which relies on accurate force feedback. Project webpage: http://prg.cs.umd.edu/FeelAnyForce