---
title: "Amazon and University of Michigan give robots a sense of touch"
description: "Robotic manipulation in contact-rich settings requires accurate tactile force perception, but existing physics simulators either run too slowly for reinforcement learning or oversimplify shear forces. HydroShear addresses this tactile reality gap by incorporating path-dependent force tracking into hydroelastic contact models. By remembering object motion history across soft elastomer sensors, the GPU-parallelizable simulator computes full 3-D contact deformations, friction, and slipping dynamics. Reinforcement learning policies trained entirely in HydroShear transferred directly to a real Franka robot equipped with GelSight Mini sensors without fine-tuning. Across four challenging manipulation tasks, HydroShear achieved a 93 percent average success rate, outperforming baseline models TacSL and FOTS."
---

# Amazon and University of Michigan give robots a sense of touch

amazon.science · Mani Nambi, Nima Fazeli · Jul 10, 2026

**Type:** Problem & solution

## Summary

Robotic manipulation in contact-rich settings requires accurate tactile force perception, but existing physics simulators either run too slowly for reinforcement learning or oversimplify shear forces. HydroShear addresses this tactile reality gap by incorporating path-dependent force tracking into hydroelastic contact models. By remembering object motion history across soft elastomer sensors, the GPU-parallelizable simulator computes full 3-D contact deformations, friction, and slipping dynamics. Reinforcement learning policies trained entirely in HydroShear transferred directly to a real Franka robot equipped with GelSight Mini sensors without fine-tuning. Across four challenging manipulation tasks, HydroShear achieved a 93 percent average success rate, outperforming baseline models TacSL and FOTS.

## Context

Robots lack scalable tactile sensing for delicate, contact-intensive tasks because simulating subtle tactile forces and shear feedback is difficult. Finite-element models are too slow for reinforcement learning, whereas faster approximations miss critical force accumulation and slipping events.

## Approach / What changed

Researchers developed HydroShear, an GPU-parallelizable simulator adding path-dependent force tracking to hydroelastic contact models. It tracks how soft elastomer membranes deform across full 3-D motion histories and is calibrated using real-world GelSight Mini sensor data isolating four physical parameters.

## Takeaways

- HydroShear enables zero-shot sim-to-real transfer of reinforcement learning policies for contact-rich robotic tasks without real-world fine-tuning.
- Across peg insertion, bin packing, book shelving, and drawer pulling, HydroShear achieved a 93% success rate, outperforming TacSL at 34% and FOTS at 58% to 61%.
- HydroShear models full 3-D motion and tracks individual surface contact points over time to compute realistic force fields that account for friction and slipping.

**Tags:** [Machine Learning](https://yomu.fyi/topic/machine-learning), [Performance](https://yomu.fyi/topic/performance)

- Source: [amazon.science](https://www.amazon.science/blog/amazon-and-university-of-michigan-give-robots-a-sense-of-touch)
- Source URL: https://www.amazon.science/blog/amazon-and-university-of-michigan-give-robots-a-sense-of-touch
- Ingested by Yomu: 2026-08-27T15:23:57.083Z

[Read original post](https://www.amazon.science/blog/amazon-and-university-of-michigan-give-robots-a-sense-of-touch)