Simultaneous Tactile-Visual Perception for Learning Multimodal Robot Manipulation

TL;DR

This paper presents TacThru, a novel sensor and learning framework that combines tactile and visual perception for improved robotic manipulation, achieving high success rates in complex real-world tasks.

Contribution

Introduction of TacThru sensor enabling simultaneous multimodal perception and TacThru-UMI framework leveraging these signals with a Transformer-based policy.

Findings

Achieved an average success rate of 85.5% on five tasks.

Outperformed tactile-only (66.3%) and vision-only (55.4%) policies.

Excelled in contact detection and precision manipulation scenarios.

Abstract

Robotic manipulation requires both rich multimodal perception and effective learning frameworks to handle complex real-world tasks. See-through-skin (STS) sensors, which combine tactile and visual perception, offer promising sensing capabilities, while modern imitation learning provides powerful tools for policy acquisition. However, existing STS designs lack simultaneous multimodal perception and suffer from unreliable tactile tracking. Furthermore, integrating these rich multimodal signals into learning-based manipulation pipelines remains an open challenge. We introduce TacThru, an STS sensor enabling simultaneous visual perception and robust tactile signal extraction, and TacThru-UMI, an imitation learning framework that leverages these multimodal signals for manipulation. Our sensor features a fully transparent elastomer, persistent illumination, novel keyline markers, and efficient tracking, while our learning system integrates these signals through a Transformer-based Diffusion Policy. Experiments on five challenging real-world tasks show that TacThru-UMI achieves an average success rate of 85.5%, significantly outperforming the baselines of tactile policy(66.3%) and vision-only policy (55.4%). The system excels in critical scenarios, including contact detection with thin and soft objects and precision manipulation requiring multimodal coordination. This work demonstrates that combining simultaneous multimodal perception with modern learning frameworks enables more precise, adaptable robotic manipulation.