Related Experiment Video
Updated: Jul 31, 2025
08:02
Robotic Sensing and Stimuli Provision for Guided Plant Growth
Published on: July 1, 2019
8.1K
Biologically inspired robotic perception-action for soft fruit harvesting in vertical growing environments
Fuli Wang1, Rodolfo Cuan Urquizo1, Penelope Roberts1
1School of Computer Science and Electronic Engineering, University of Essex, Colchester, CO4 3SQ UK.
Precision Agriculture
|May 8, 2023
Summary
A new robotic system uses biomimicry and synthetic data to harvest strawberries, addressing labor shortages in agriculture. This automated solution efficiently identifies and picks ripe fruit in vertical farms.
Area of Science:
- Agricultural Robotics
- Biomimetic Systems
- Computer Vision
Background:
- Global labor shortages are impacting physically demanding agricultural tasks, such as soft fruit harvesting.
- Existing harvesting methods face challenges with labor availability due to demographic, migration, and economic factors.
- Vertical farming environments present unique challenges for automated harvesting systems.
Purpose of the Study:
- To present a biomimetic robotic solution for the complete 'Perception-Action' loop in strawberry harvesting.
- To address crop and environmental variability in harvesting operations.
- To develop a robot action system capable of handling runtime task constraints.
Main Methods:
- Utilized conditional Generative Adversarial Networks (GANs) with synthetic data for ripe fruit identification, avoiding manual data collection and labeling.
- Employed image-to-image translation for effective synthetic data training.
- Integrated stereo camera-generated point cloud data for fruit localization and a Passive Motion Paradigm framework for robotic arm control, inspired by neural movement control.
Main Results:
- Successfully demonstrated strawberry detection, stem reaching, and cutting in field trials.
- Extended the system's capabilities to analyze complex canopy structures and enable bimanual coordination for searching and picking.
- Achieved efficient fruit identification using synthetic data, overcoming limitations of traditional deep neural networks.
Conclusions:
- The proposed biomimetic robotic system offers a viable solution to labor shortages in soft fruit harvesting.
- The perception system effectively uses synthetic data, reducing the need for extensive real-world data collection.
- The action system, guided by the Passive Motion Paradigm, shows promise for adaptable and efficient robotic harvesting, with potential for adaptation to other crops like tomatoes and peppers.
