Prevention of Frost Damage to Apple Tree Bud Stages Using an Autonomous Multi-Vehicle System

Open Access
- Author:
- Canto, Maxfield
- Graduate Program:
- Agricultural and Biological Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- June 13, 2022
- Committee Members:
- Suat Irmak, Program Head/Chair
David James Lyons, Thesis Advisor/Co-Advisor
Paul Heinz Heinemann, Committee Member
Long He, Committee Member
Henry Joseph Sommer, III, Committee Member
Daeun Choi, Special Signatory - Keywords:
- precision agriculture
agriculture
robotics
sensing
frost
frost damage
UAV
UGV
autonomy
autonomous vehicles
apple buds
apple orchard
cyber-physical system
IoT
ROS
thermal mapping
RTK GPS - Abstract:
- Fruit growers have dealt with crop loss due to extreme weather events since the first orchards were cultivated. In the United States today, frost damage alone leads to more economic loss than any other weather-related occurrence. Despite this, the temperature monitoring and frost prevention techniques used by many operations to this day are crude, ineffective, and labor intensive. The goal of this research was to develop and test a multi-vehicle cyber-physical system that can monitor and react to frost in an orchard environment without the need for human intervention. The objectives of the study were to: (1) develop an aerial vehicle and vision algorithm that can locate regions of high frost risk using thermal imaging, (2) design a ground vehicle and algorithm that can plan and execute the most favorable route to areas of concern in a test orchard, and (3) create a wireless network for real-time communication between the vehicles and a ground control station. The final system used a UAV and FLIR imaging to monitor a test orchard. The vision algorithm was developed to georeference the thermal images and locate cooled ground truth targets. The path planning algorithm then designed routes and navigated the ground vehicle to the target regions in both simulated and field tests. A ground vehicle was retrofitted with a high-accuracy GPS receiver for reliable autonomous navigation in the orchard environment. Communication between the vehicles was handled through a purpose-built ROS© framework. In the field, the UAV’s imaging system could locate ground truth targets within 4.6 ± 3.6 m and 4.8 ± 2.9 m in the lateral and longitudinal axes respectively. In simulation, the path planning algorithm could determine the optimal routes to target locations in 85% of the test cases. During further field testing, the ground vehicle successfully navigated to all target locations with minimal path deviations of 3.31 and 3.10 cm in the lateral and longitudinal axes.