Autonomous Agricultural Spraying Project

Project Team

Herman Herman -- Project Manager
Dave Stager -- Lead Engineer
Carl Wellington -- Graduate Student
Cris Dima -- Graduate Student
Bob McCall -- Research Programmer
Tony Stentz -- Principal Investigator

Many thanks to past project members: Mark Ollis, Chris Fromme, Frank Campagne, and Jim Ketterer.

Project Description

[Spraying in a Florida grove]

The spraying of groves, orchards and row crops such as citrus, apples, tomatoes, and corn with pesticides is a vital operation in the agriculture industry. Pesticides are applied to crops anywhere from a few times a year to a few times a week. Additional applications are needed when an infestation is detected, and the pesticides must be applied rapidly. Spraying from the air is a fast way to deploy the compounds, but it is costly, wasteful, and detrimental to the environment, since it blankets the entire field. Furthermore, spraying groves from the air is ineffective due to poor penetration of the canopy. Spraying from the ground is more efficient, since the sprayers can target individual trees or crow rows, but it is much slower. Spraying is most effective when performed at night, since insects are most active at that time and there is no sunlight to breakdown the pesticide compounds. Night spraying is also safer since there are fewer workers in the field to be exposed to the chemicals. The problem with night spraying is that poor visibility and fatigue limit the operator's productivity and increases the chance of accident and injury.

The objective of this project is to automate ground-based vehicles for pesticide application, such that a single operator, from a remote location, can oversee the operation of at least four spraying vehicles running at night. The potential savings are immense. There are over 6B application-acres of vegetables and row crops in the U.S. per year. If only 50 cents per acre were saved due to automatic spraying of 3% of the crop (the likely percentage of innovators willing to try new technology), the net savings to the industry would be $90M/year. As the technology is widely accepted, the savings would be in the billions.

The project sponsors are NASA, USDA Agricultural Research Service, and Deere & Company. We also thank Dean Remick of Eden BioScience Corporation for his promotion of the project, and U.S. Sugar Corporation and A. Duda & Sons for access to their groves.

Automation will enable:

Each worker employed in spraying to be up to four times more productive, by overseeing a fleet of spraying vehicles rather than driving just one;
Each spraying rig to be at least 20% more productive, due to constant operation at the highest, safe speed;
Spraying to be performed at night, when the compounds are most effective and the proper time and care can be taken to ensure complete coverage;
Compounds with higher efficacy to be used, since human exposure will be minimized.

In addition to the cost savings, precision application of the chemicals will minimize environmental damage. The first year of the program is focussing on tree crops, such as citrus and apples. The second year of the program will focus on vegetables and other row crops.

Technical Objectives

The technical objectives are:

Produce a primary navigation system based on GPS and dead reckoning sensors with an accuracy of +/- 6 inches or better.
Produce a second navigation system based on camera vision or ladar to serve as a backup in the event of GPS unavailability or inaccuracies. The secondary system should be accurate as the primary for short distances.
Produce an obstacle detection system with zero false negatives. This will be accomplished by taking advantage of domain specific knowledge. False positives will be handled via supervisor confirmation.
Produce an operator control system that enables a supervisor to track the locations of the sprayers, confirm obstacles, and remotely drive the machines such that the supervisor minimally delays the overall operation.
Produce an integrated spraying system capable of operating evenly, at the highest possible rate, spraying chemicals only as needed. We expect to achieve a 20% performance increase per machine across a shift, and a 20% reduction in the amount of chemical applied per rig.

Project Status

In the last week of April, 2000, we transported our Deere 6410 tractor to an orange grove in Florida to collect data and test the autonomous navigation system for trees. For one of the tests, the system was taught to drive a 7 kilometer path through the grove. It was then put in autonomous mode and drove the taught path at an average speed of 4 mph, spraying water enroute. The pictures below show the unmanned tractor (front and rear) driving down a row.

[Spraying in a Florida grove]

The pictures below show a sequence of tractor shots as it executed a turn. The camera was mounted on a tower near the grove.

[Spraying in a Florida grove]

For more information, contact:

Tony Stentz
National Robotics Engineering Consortium
#10 Fortieth Street
Pittsburgh, PA 15201
Ph: (412) 268-8155
Fax: (412) 681-6961
Email: tony+@cmu.edu

Back to home page.