MidasTouch: Monte-Carlo inference over distributions across sliding touch

TL;DR

MidasTouch introduces a Monte-Carlo based tactile perception system for real-time localization of a sliding sensor on object surfaces, leveraging learned surface geometry representations without visual priors.

Contribution

It develops a novel tactile localization framework using a Monte-Carlo particle filter and a learned surface geometry code network, enabling online pose estimation without visual cues.

Findings

Effective localization on YCB objects demonstrated

Rapid disambiguation of sensor pose through surface geometry traversal

Release of a new dataset for tactile-object interaction

Abstract

We present MidasTouch, a tactile perception system for online global localization of a vision-based touch sensor sliding on an object surface. This framework takes in posed tactile images over time, and outputs an evolving distribution of sensor pose on the object's surface, without the need for visual priors. Our key insight is to estimate local surface geometry with tactile sensing, learn a compact representation for it, and disambiguate these signals over a long time horizon. The backbone of MidasTouch is a Monte-Carlo particle filter, with a measurement model based on a tactile code network learned from tactile simulation. This network, inspired by LIDAR place recognition, compactly summarizes local surface geometries. These generated codes are efficiently compared against a precomputed tactile codebook per-object, to update the pose distribution. We further release the YCB-Slide dataset of real-world and simulated forceful sliding interactions between a vision-based tactile sensor and standard YCB objects. While single-touch localization can be inherently ambiguous, we can quickly localize our sensor by traversing salient surface geometries. Project page: https://suddhu.github.io/midastouch-tactile/