Researchers Leverage Agbot for In-Crop Sampling and Monitoring Using Unmanned Ground Vehicles
Outcome/Accomplishment
Monitoring crop health is essential for successful production, but it has been a labor-intensive process that can impede other important operations on the farm. Existing technologies are also unlikely to adequately characterize actual field variability. In response, researchers have developed a system that monitors important crop characteristics using unmanned ground vehicles (UGVs) and a system of sensors. Recent work has focused on the development of algorithms that are capable of fine sensor measurements for navigation as well as physical sampling. This work is supported by the National Science Foundation (NSF)-funded Internet of Things for Precision Agriculture (NSF IoT4Ag) Engineering Research Center (ERC), which is headquartered at the University of Pennsylvania (Penn) with Purdue University as a partner institution.
Impact/Benefits
Precision farming relies on automation and technology that provide timely data for better decision-making. The NSF IoT4Ag system overcomes the challenges of existing robotic platforms: those designed to move between rows often cannot navigate under plant canopies, while larger systems that travel above crops can be too heavy, frequently leading to soil compaction. The NSF IoT4Ag system enables in-row and under-canopy navigation in corn fields, collects physical samples, and integrates a suite of sensors to measure soil and plant conditions—allowing farmers to take corrective measures more quickly and effectively.
Explanation/Background
The NSF IoT4Ag system employs UGVs outfitted with systems that use LiDAR (Light Detection and Ranging) technology to simultaneously locate and map corn fields, model predictive control, and receive data from soil and plant tissue sampling. These capabilities provide a scalable approach to monitoring and managing crop health.
The research team based at Purdue and Penn developed a series of algorithms to guide the work of the UGVs. P-AgSLAM is a LiDAR-based algorithm for simultaneous localization and mapping in corn fields. It has been validated in simulation and in the testbed. Another algorithm measures the distance from the ground to the point where the corn ear attaches to the stalk; it produced less than 5% errors when tested on tall and short corn varieties. P-AgNAV is a LiDAR-based navigation algorithm that uses range-view images and models predictive control. It was validated in the ACRE testbed along with buried soil sensor deployment and sensor reading modules designed for P-AgBots and a Polaris Range UTV trailer. Deep Learning-based Leaf Detection for Robotic Physical Sampling was also evolved using Mask R-CNN and ROLOv5. The transformer-based AI approach for detection achieved 6DoF (degrees of freedom) pose of corn leaves for physical sampling and offered a success rate greater than 90%. The research team includes Drs. David Cappalleri, Melba Crawford, Daniel Quinn, and James Krogmeier (Purdue) as well as Vijay Kumar (Penn).
Location
Philadelphia, Pennsylvaniawebsite
Start Year
Microelectronics and IT
Quantum, Microelectronics, Sensing, and IT
Lead Institution
Core Partners
Fact Sheet
Outcome/Accomplishment
Monitoring crop health is essential for successful production, but it has been a labor-intensive process that can impede other important operations on the farm. Existing technologies are also unlikely to adequately characterize actual field variability. In response, researchers have developed a system that monitors important crop characteristics using unmanned ground vehicles (UGVs) and a system of sensors. Recent work has focused on the development of algorithms that are capable of fine sensor measurements for navigation as well as physical sampling. This work is supported by the National Science Foundation (NSF)-funded Internet of Things for Precision Agriculture (NSF IoT4Ag) Engineering Research Center (ERC), which is headquartered at the University of Pennsylvania (Penn) with Purdue University as a partner institution.
Location
Philadelphia, Pennsylvaniawebsite
Start Year
Microelectronics and IT
Quantum, Microelectronics, Sensing, and IT
Lead Institution
Core Partners
Fact Sheet
Impact/benefits
Precision farming relies on automation and technology that provide timely data for better decision-making. The NSF IoT4Ag system overcomes the challenges of existing robotic platforms: those designed to move between rows often cannot navigate under plant canopies, while larger systems that travel above crops can be too heavy, frequently leading to soil compaction. The NSF IoT4Ag system enables in-row and under-canopy navigation in corn fields, collects physical samples, and integrates a suite of sensors to measure soil and plant conditions—allowing farmers to take corrective measures more quickly and effectively.
Explanation/Background
The NSF IoT4Ag system employs UGVs outfitted with systems that use LiDAR (Light Detection and Ranging) technology to simultaneously locate and map corn fields, model predictive control, and receive data from soil and plant tissue sampling. These capabilities provide a scalable approach to monitoring and managing crop health.
The research team based at Purdue and Penn developed a series of algorithms to guide the work of the UGVs. P-AgSLAM is a LiDAR-based algorithm for simultaneous localization and mapping in corn fields. It has been validated in simulation and in the testbed. Another algorithm measures the distance from the ground to the point where the corn ear attaches to the stalk; it produced less than 5% errors when tested on tall and short corn varieties. P-AgNAV is a LiDAR-based navigation algorithm that uses range-view images and models predictive control. It was validated in the ACRE testbed along with buried soil sensor deployment and sensor reading modules designed for P-AgBots and a Polaris Range UTV trailer. Deep Learning-based Leaf Detection for Robotic Physical Sampling was also evolved using Mask R-CNN and ROLOv5. The transformer-based AI approach for detection achieved 6DoF (degrees of freedom) pose of corn leaves for physical sampling and offered a success rate greater than 90%. The research team includes Drs. David Cappalleri, Melba Crawford, Daniel Quinn, and James Krogmeier (Purdue) as well as Vijay Kumar (Penn).