Source: The Regents of University of California submitted to
VISION-GUIDED FLEXIBLE ROBOTS FOR HARVESTING THE INTERIOR OF TREE CROPS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
NEW
Funding Source
Reporting Frequency
Annual
Accession No.
1031001
Grant No.
2023-67021-40629
Project No.
CALW-2022-11471
Proposal No.
2022-11471
Multistate No.
(N/A)
Program Code
A1521
Project Start Date
Sep 1, 2023
Project End Date
Aug 31, 2025
Grant Year
2023
Project Director
Sheng, J.
Recipient Organization
The Regents of University of California
200 University Office Building
Riverside,CA 92521
Performing Department
(N/A)
Non Technical Summary
This project is motivated by an urgent need for an autonomous system to enable more sustainable agricultural production. Agricultural production is heavily labor-intensive; yet, the increase of labor costs, impact of COVID-19 pandemic, and decrease of next generation's interest in farm labor are making it increasingly difficult to find sufficient labor. Hence, automation techniques are sought to perform agricultural tasks in a timely manner. However, most progress in agricultural automation took place in relatively simple zones that are close to the ground and harvesting tree crops remains the biggest challenge due to large and unstructured canopies and rough terrains. Robotic technologies are promising to address this challenge due to their adaptability. However, their harvesting efficiency is controversial due to the inability to harvest inside the canopy.In this seed project, we will focus on 1) developing a novel soft continuum robot that can navigate in cluttered environments, integrated with a harvesting end effector, 2) developing a vision system with an array of on-board cameras that can detect crops inside the canopy, and 3) evaluating the performance of the robot system by harvesting mandarin oranges in the lab setting and then on real trees in the orchard. Through the methods mentioned above, this seed project will generate convincing preliminary data to address the feasibility of robotically harvesting the Interior of Tree Crops and support our long-term objective of autonomous robotic harvesting in collaboration with traditional robots that are efficient to harvest surface crops. The project will benefit profitability and sustainability of the tree crop industry by mitigating labor shortage, automating harvesting, and preserving the integrity of tree canopies, and ultimately enhance the quality of life for farmers and society.
Animal Health Component
0%
Research Effort Categories
Basic
0%
Applied
50%
Developmental
50%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
50106992020100%
Goals / Objectives
The goal of this project is to address the technical hurdle on how to harvest crops inside the tree canopywith the following three objectives: 1) Design and develop a novel flexible robot with cameras and necessary harvesting accessories, 2) Design and develop computer vision models to detect and localize harvesting targets and obstacles, and 3) Assess our to be developed technologies in engineered and real-life harvesting environments.
Project Methods
A novel robot design will be implemented systematically, and a new robot navigation system will be created by applying cutting-edge computer vision and machine intelligence. We will expand our preliminary work and develop a robot equipped with harvesting accessories, a mobile platform, and cameras. The robot will be designed to consist of multiple telescopic cable-driven continuum structures. Unlike traditional agricultural robots, this new robot design will enable follow-the-leader navigation in unstructured environments. Deep learning methods will be used to reconstruct 3D canopy structures and identify and track branches and fruits using camera images. We will first use the cameras on the side and bottom of the tree canopy to capture multiple images along a fixed trajectory, which we will use to recover the initial 3D structure of the tree canopy. We will then continuously update the internal canopy structure using images captured by the cameras mounted on the manipulator and detect major branches and fruits. After testing the proposed technologies in the lab setting to verify the functionality of each module, we will evaluate the potential of the integrated system for avocado trees with moderate-size canopies using UCR Agricultural Operations facilities and measure the detection rate, cycle time, and damage rate of robotic harvesting.