Design of an Autonomous Ag Platform Capable of Field Scale Data Collection in Support of Artificial Intelligence
1A. Balmos, 1S. Jha, 2J. Krogmeier, 2D. Buckmaster, 3D. J. Love, 4R. H. Grant, 5M. Crawford, 3C. Brinton, 3C. Wang, 6D. Cappelleri
1.
2. Purdue University
3. Electrical and Computer Engineering, Purdue University
4. Department of Agronomy, Purdue University
5. School of Civil Engineering, Purdue University
6. School of Mechanical Engineering, Purdue University
The Pivot+ Array is intended to serve as an innovative, multi-user research platform dedicated to the autonomous monitoring, analysis, and manipulation of crops and inputs at the plant scale, covering extensive areas. It will effectively address many constraints that have historically limited large-scale agricultural sensor and robotic research. This achievement will be made possible by augmenting the well-established center pivot technology, known for its autonomy, with robust power infrastructure, high-speed fiber and wireless networking capabilities. The system will also include environmentally controlled and protected cabinets, housing non-field-hardened sensors, data servers, and signal boosters, which enable communications for remote data access.
As a result, the Pivot+ Array will offer a collaborative, field-scale research platform, focusing on agricultural sensors, sensor communications, robotics, and machine learning applications, including micro-meteorology, phenotyping, and crop systems management. The platform's design will encompass an expansive 115-acre interaction network, facilitating the execution of numerous parallel experiments. Researchers will have the flexibility to design field trials within sectors or adapt plots to varying sizes and shapes within the circular span. Data monitoring will occur at regular 90-minute intervals, enabled by the pivot's ability to operate in diverse weather and lighting conditions. This capability will allow for comprehensive sensing of soil-water-plant-environment interactions, a feat unattainable with current instruments but crucial for AI research.
Moreover, researchers can leverage high-resolution and gridded data, including soil order, drainage class, elevation, slope, and topographic position classifiers, to tailor site-specific experiments. The outcome will be a wealth of high-resolution temporal and spatial observations and control over field trials, yielding the high-quality datasets demanded by modern AI and machine learning methodologies. The paper will cover the design choices made in the Pivot+ Array along with a case study to showcase its use in agronomic field experiments.