The Role of Tactile Sensing for Learning Reach and Grasp

TL;DR

This paper investigates how different tactile sensing configurations influence reinforcement learning for robotic grasping, highlighting that tactile features can improve learning under visual perception errors but complex inputs may hinder training.

Contribution

It provides a systematic comparison of tactile sensing setups in reinforcement learning for grasping, addressing the impact of tactile complexity on learning performance.

Findings

Tactile features improve grasp learning under visual perception errors.

Complex tactile inputs can complicate the training process.

Different tactile setups have varying effects on learning outcomes.

Abstract

Stable and robust robotic grasping is essential for current and future robot applications. In recent works, the use of large datasets and supervised learning has enhanced speed and precision in antipodal grasping. However, these methods struggle with perception and calibration errors due to large planning horizons. To obtain more robust and reactive grasping motions, leveraging reinforcement learning combined with tactile sensing is a promising direction. Yet, there is no systematic evaluation of how the complexity of force-based tactile sensing affects the learning behavior for grasping tasks. This paper compares various tactile and environmental setups using two model-free reinforcement learning approaches for antipodal grasping. Our findings suggest that under imperfect visual perception, various tactile features improve learning outcomes, while complex tactile inputs complicate training.