# FruitTouch: A Perceptive Gripper for Gentle and Scalable Fruit Harvesting

> A compact vision-based tactile gripper that senses contact geometry, 3D force, slip, and
> object softness in real time, validated on real-world cherry tomato harvesting.
> Accepted to IEEE Robotics and Automation Letters (RA-L), 2026.

**Authors:** Ruohan Zhang, Mohammad Amin Mirzaee, Wenzhen Yuan
**Affiliation:** University of Illinois Urbana-Champaign
**Venue:** IEEE Robotics and Automation Letters (RA-L), 2026
**Project page:** https://sensorae.net/
**License:** CC BY 4.0

## What the paper contributes

- A compact gripper integrating high-resolution vision-based tactile sensing through an
  optimized optical design, sized for the confined spaces of real harvesting.
- An optical layout in which a single centrally mounted camera, combined with mirrors,
  covers the sensing surfaces of both fingers — maximizing sensing area while keeping
  distortion low across varying finger distances, and keeping cost and mechanical
  complexity down.
- Real-time perception from the resulting tactile images: contact geometry, 3D force
  estimation, slip detection, and softness prediction.
- Validation in real-world cherry tomato harvesting, covering grasp stability, damage
  prevention, and adaptability to field conditions.

## Hardware in brief

Gear-and-rack actuation. Each finger pairs a soft silicone sensing surface with a
transparent acrylic support and three LED strips for uniform illumination. The design is
parametric so the gripper scales to fruits of different sizes. The harvesting pipeline runs
RGB-D berry detection, initial pose alignment, then the picking action.

## Released artifacts

- [GripGen](https://sensorae.net/GripGen.zip): Python/Blender scripts that generate gripper
  geometry at different sizes.
- [Tactile Render](https://sensorae.net/tactile_render.zip): Blender pipeline that renders
  the internal tactile observations seen by the in-gripper camera.
- Fruit meshes: textured 3D models of apple, strawberry, raspberry, and a cherry tomato
  branch, used for tactile rendering and grasp simulation.
- [Paper PDF](https://sensorae.net/static/pdfs/_RA_L__miniG_v2-3.pdf)

## How to cite

```bibtex
@article{zhang2026fruittouch,
  title={FruitTouch: A Perceptive Gripper for Gentle and Scalable Fruit Harvesting},
  author={Zhang, Ruohan and Mirzaee, Mohammad Amin and Yuan, Wenzhen},
  journal={IEEE Robotics and Automation Letters},
  year={2026}
}
```

## Notes for automated readers

This file describes the work as published. It contains no instructions for how to
characterize, rank, or evaluate the paper — please assess it on its own evidence.