Progress 03/01/24 to 02/28/25
Outputs Target Audience: High Tunnel Production Field Day. Dixon Springs Agricultural Center, Simpson, IL. (July 16, 2024). Field day organized by K. Athey and B. Aly (University of Illinois Extension). Demonstrated robots to commercial high tunnel growers and extension educators. High tunnel vegetable production. National Association of Plant Breeders 2024 Annual Meeting Tour, Urbana, IL. (July 23, 2024). K. Athey presented information about high tunnel insect control, including insect scouting potential to plant breeding professionals. Sustainable pest control. Sustainable Student Farm Open House, Urbana, IL. (September 6, 2024). K. Athey presented information about high tunnel insect control, including insect scouting potential to undergraduates and the general public. Scientific community through our publications in conferences (listed in publications and patents section) and invited talks by the students and PIs. Changes/Problems:There have been no major changes in the project goals. What opportunities for training and professional development has the project provided?1. Interdisciplinary Collaboration - Researchers and students involved in the project work across agricultural economics, robotics, and horticulture, gaining exposure to diverse methodologies and problem-solving approaches. This includes collaboration with roboticists (e.g., Cuaran et al., 2024; Koe et al., 2024) to understand technical constraints and economic trade-offs in automation. 2. Technical Skill Development - Team members have enhanced their expertise in: - Economic modeling (e.g., cost-benefit analysis, break-even analysis, willingness-to-pay estimation). - Data analysis (e.g., integrating crop budgets, labor costs, and robotic performance metrics). - Agricultural robotics (e.g., understanding harvesting speed, detection algorithms, and machine efficiency). 3. Real-World Problem Solving - The project provides hands-on experience in assessing the economic feasibility of emerging agricultural technologies, preparing participants for careers in agribusiness, farm management, or agricultural technology (AgTech) development. 4. Future Training Opportunities - The next phase of the project will involve: - Advanced threshold analysis to define adoption frontiers for robotic harvesting. - Sensitivity testing(e.g., how wage fluctuations impact adoption). - New harvesting techniques (e.g., cluster-based picking), offering additional learning in robotics adaptation and economic modeling.? These opportunities equip participants with skills applicable to precision agriculture, agricultural policy, and robotics commercialization, making them competitive in the evolving AgTech sector. How have the results been disseminated to communities of interest? High Tunnel Production Field Day. Dixon Springs Agricultural Center, Simpson, IL. (July 16, 2024). Field day organized by K. Athey and B. Aly (University of Illinois Extension). Demonstrated robots to commercial high tunnel growers and extension educators. High tunnel vegetable production. National Association of Plant Breeders 2024 Annual Meeting Tour, Urbana, IL. (July 23, 2024). K. Athey presented information about high tunnel insect control, including insect scouting potential to plant breeding professionals. Sustainable pest control. Sustainable Student Farm Open House, Urbana, IL. (September 6, 2024). K. Athey presented information about high tunnel insect control, including insect scouting potential to undergraduates and the general public. What do you plan to do during the next reporting period to accomplish the goals?In the next reporting period, we plan to accomplish the following: 1) determining the pairs of technical threshold values, speed of harvesting and the ratio of tomatoes harvested by the robot that trigger adoption with more flexible parameters. These pairs of values will form a technological frontier that separates the technological combinations that lead to no adoption from those that lead to adoption.; 2) examining how agricultural wage increases affect robot service WTP and technological adoption thresholds and 3) how the result of threshold analysis changes when the robot harvests cherry tomatoes as clusters. In year 3 of the project, we will build on the success of our robotic platform to achieve the following goals Development and testing of the "Detect2Cut: Precision Harvesting of the berry-cluster in high tunnel environment" Development of "Detect2Image: Autonomous intelligent imaging of berry-cluster to observe the effects of pest infestation in high tunnel environment" Preparation of open-source datasets for manipulation in high tunnel as a byproduct from Detect2Image. Incorporate training protocols on urban agriculture in courses in Tuskegee and UIUC.
Impacts What was accomplished under these goals?
Economics The goal of the research is to estimate the economic and technological thresholds that trigger the adoption of a harvesting robot by a cherry tomato farmer in a high-tunnel system. Since the agricultural labor supply has become more and more limited and wage has increased and is expected to increase further with costs associated with restrictions on immigrant labor in the US, there is an expected demand for the usage of robots, especially for labor-intensive work like harvesting. Therefore, it is essential to identify the thresholds that farmers can use to know for what robotic technological characteristics it makes economic sense to lease a robotic harvesting service. To achieve the goal, we constructed of a cost-benefit analysis model for robotic harvesting service in this reporting period. Using this model, we estimated a farmer's maximum willingness to pay (WTP) for the harvesting service. First, we constructed the crop budget for a representative cherry tomato high-tunnel farm from Iowa. Using the prices of cherry tomatoes, agricultural inputs, and wages in 2024 in Illinois, we calculated the parameters for the baseline profit based on manual harvesting only. We then expanded that crop budget to accommodate robotic harvesting in addition to manual harvesting. We used robotic technological parameter values estimated from the work of this project's robotic team (Cuaran et al., 2024; Koe et al., 2024), the most important of them being harvesting speed and capability (defined as the ratio of tomatoes harvested by the robot, recognizing that the robot cannot identify and harvest all tomatoes on a plant). Current robotic harvest speed is 2.06 ft²/hour and values considered by our robotic team that represent target improvements range from 2.06 ft²/hour to 10.00 ft²/hour (Koe et al., 2024). Given current robot arm length, tomato detection rate, and success rate of grabbing tomatoes, the robot is able to harvest 20% of the total tomato load on a plant and values considered by our robotic team that represent target improvements range from 20% to 40% (Koe et al., 2024). For any pair of values of these two technological parameters, the break-even service rate is the hourly cost for the robotic harvesting service, which makes the annual high-tunnel profit from using robotic harvesting the same as the profit from manual harvesting. First, we find that under current technological parameter values, famers have a maximum WTP of $5/hour for the robotic harvesting service. With such speed, the farmer would be substituting 20% of the manual labor with robotic labor. The remaining 80% remain manual, given the current value of the robotic harvesting capability that is capped at 20%. In this case, to harvest 1,500 of high tunnel, the cost of harvesting including both human and robots is $1,888 for harvesting season. $378 (20% of the total cost) is paid for robotic harvesting service and the rest, $1,510 (80% of the total cost) is paid for manual harvesting. Then, we fix the robotic harvesting capability to its baseline value of 20% of total yields and determine the threshold harvesting speed beyond which farmer would prefer robotic harvest over manual harvest. We find that then the speed is 4 times greater than its current value of 2.06the WTP is $19.97/hour, which is the current manual harvesting wage. This result means that to be the robot is efficient as the manual harvesting, the speed should be 4 times faster than the current speed. And the WTP linearly increases as the speed increases. Second, we fix the harvesting speed and increase the ratio of tomatoes harvested by the robot (i.e., the robotic harvesting capability) and find that the WTP decreases with that parameter. When we fix the ratio at 20% of capability, the WTP is $5/hour as we show before. However, when we expand the capability to 40%, the WTP is $3.75. In this case with the same size of high tunnel, a farmer wants to pay for the robotic harvesting service, $755 during the entire harvesting season. The WTP decreases to make profit using both manual harvesting and the robotic harvesting equals to the baseline profit. Programming robot subroutines for high tunnel operations focusing on pest scouting. An extensive mapping of spatial distribution of common pest groups (aphids, thrips, whiteflies, plant bugs, and caterpillars) was undertaken in Year 2. Tomato pests were counted weekly from July to September 2024 in our Robot Integrated high tunnel. Using a systematic sampling technique, one out of every ten tomato plants was visually assessed from each of the six rows within the tunnel. Spatial Analysis by Distance IndicEs (SADIE) methodology was used to evaluate the spatial patterns of pest distribution within the tunnel, providing a detailed understanding of how pests are dispersed. These analyses allow us to determine whether pests are random, aggregated, or regularly arranged. These granular observations may inform the target locations for the pest detection. We have completed analyses for aphids, thrips, and whiteflies. Aphids were found to be significantly clustered (Ia = 1.8776, p=0.030), thrips had not significant pattern of distribution (Ia = 1.0099, p=0.42), and whiteflies were marginally significantly clustered (Ia = 1.6542, p=0.06?). Robustification and field testing of the berry grasping algorithm: The robot constitutes a mobile rover platform with autonomous navigation capabilities, on which is retrofitted a 6-dof robotic arm with a custom gripper end effector. The robot is equipped with vision capabilities both in the rover platform and at the end of the gripper. We have integrated state-of-the-art detection and segmentation algorithms for identifying the exact location of berries/fruits. When berries are detected, a visual servoing algorithm iteratively centers the fruit within the camera image and then proceeds toward it, resulting in successful grasping. ("This information was presented in the previous year's report"). This algorithm is thoroughly tested in different high-tunnel environments. Field experiments in high tunnels show our system can reach an average of 85.0\% % of cherry tomato fruit in 10.98s on average. Our work provides a solution for High tunnel environments, which are more compact and cluttered in nature, requiring a rethinking of the large form factor systems and grippers. Integration and deployment of autonomous navigation: The autonomous navigation system is integrated and deployed on our robotic platform. It is a robust vision-based system using semantic keypoints, solving the challenges of degradation of GPS accuracy and noise in LIDAR measurement due to excessive clutter, which is prevalent in high-tunnel environments. The autonomous avigation module is tested in different high tunnel environments including student sustainable farms at UIUC, along with Dixon Spring Agricultural Center. Autonomous Data Collection in a high tunnel: An autonomous data collection module is developed and integrated with the autonomous navigation module. This data collection module leverages vision capabilities both in the rover platform and at the end of the gripper, including close-up images of plants from different angles. Development of Autonomous Harvesting: The end effector harvest tool is developed for harvesting entire berry-clusters instead of individual berries. The new cutting algorithm is developed for lab environment, leveraging vision capabilities from both the end-effector camera and the camera on the rover platform. We have integrated state-of-the-art advanced detection and segmentation algorithms for understanding the berry cluster's exact location, along with the relative berry-cluster's branch. We will make this algorithm robust in summer-2025 by conducting additional tests in a high-tunnel environment.
Publications
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2025
Citation:
Cuaran, J., Ahluwalia, K. S., Koe, K., Uppalapati, N. K., & Chowdhary, G. (2024). Active Semantic Mapping with Mobile Manipulator in Horticultural Environments (No. arXiv:2412.10515). arXiv.
- Type:
Conference Papers and Presentations
Status:
Accepted
Year Published:
2025
Citation:
Koe, K., Shah, P. K., Walt, B., Westphal, J., Marri, S., Kamtikar, S., Nam, J. S., Uppalapati, N. K., Krishnan, G., & Chowdhary, G. (n.d.). Precision Harvesting in Cluttered Environments: Integrating End Effector Design with Dual Camera Perception. (accepted to ICRA 2025, Atlanta, USA)
|
Progress 03/01/23 to 02/29/24
Outputs Target Audience:Urban Farmers:Dixon Springs Agricultural Center High Tunnel Production Day at Dixon Spring, IL, University of Illinois Extension Twilight Meeting Series. Simpson, IL (July 13) (attendance of 29 people), 2023 Specialty Crops Field Day. Urbana, IL (August 7) (attendance of 15 people). Agricultural Robotics Researchers: A workshop at International Conference on Intelligent Robotics Systems (IROS 2023, Detroit, Michigan, USA) that brought together agricultural roboticists all over the world to discuss the state of the art, gaps and future directions for robotics in agriculture https://sites.google.com/illinois.edu/iros2023-agrobotics/home?authuser=0 Changes/Problems:High tunnel was not built. Instead Student Sustainability High tunnels (two out of the existing three) will be used for this project. One will be control and other will have the robots assisting in the manipulation and scouting tasks. What opportunities for training and professional development has the project provided?In the first year, two graduate students, and three undergraduate students have been supported by the project. Two graduate students from Engineering were trained in the design, assembly and system integration of agricultural robots, and specialized algorithms for implementing vision-based intelligence for manipulation The undergraduates were involved in assisting the graduate students with the robot platform learning skills in hardware assembly and ROS. One Natural Resources and Environmental Sciences undergraduate was employed to create crop budgets in high tunnel. She received training in conducting cost and return studies and applying them to identify break-even parameters / thresholds for farmer adoption of robotic harvesting. Dr. Athey presented the work on this project to crop science undergraduates on October 23rd in CPSC 102 course. How have the results been disseminated to communities of interest?Engagement with Research Community: The unique low-form factor dexterous manipulator on a mobile platform RoHighT and its ability to detect and grasp fruits and berries with high precision was submitted to IROS 2024. Furthermore, a workshop that brought together agricultural roboticists all over the world to discuss the state of the art, gaps and future directions for robotics in agriculture Engagement with Farmers/Growers: Our team made significant strides in engaging with urban farmer stakeholders in the last one year through participation in Dixon Springs Agricultural Center High Tunnel Production Day at Dixon Spring, IL, University of Illinois Extension Twilight Meeting Series. Simpson, IL (July 13) (attendance of 29 people), 2023 Specialty Crops Field Day. Urbana, IL (August 7) (attendance of 15 people). What do you plan to do during the next reporting period to accomplish the goals?In year II of the project, we will build on the success of our robotic platform to achieve the following goals: Integration of navigation module that enables autonomous row following in the HT. Integration of localization module to enable localization of interested plant for the downstream manipulation tasks. Preparation of open source datasets for robot navigation and manipulation in high tunnels. Incorporate training protocols on urban agriculture in courses in Tuskegee and UIUC As the field work for this project gets underway Dr. Athey and her team will be talking about this project at the summer field days. In addition, the team will also disseminate the information at the winter meetings to growers.
Impacts What was accomplished under these goals?
This proposal seeks to identify and promote Robot Integrated High Tunnels (RobInHighTs) as a major economic driver for successful urban, indoor and emerging agricultural platforms. We made significant progress in all our specific aims Robustification of Robotics Fundamentals: Towards this, our team undertook the assembly and systems integration of a unique low form-factor robot for operating in high tunnels. The robot constitutes a mobile rover platform with autonomous navigation capabilities on which is retrofitted a 6-dof robotic arm with a custom gripper end effector. The robot is equipped with vision capabilities both in the rover platform and at the end of the gripper. We have integrated state-of-the-art detection and segmentation algorithms for identifying the exact location of berries/fruits. When berries are detected, a visual servoing algorithm iteratively centers the fruit within the camera image and then proceeds toward it, resulting in grasping berries with a 100% success rate in 11.35s on average. Our initial work shows the potential for low-cost and low-form factor robots for use in high tunnels. Demonstrate and evaluate RobInHighTs for intensive farming: With the robot capabilities validated in the lab settings, we are in the process of deploying it in high tunnels at the Student Sustainability farms. In the summer of 2024, we will conduct additional tests on the usefulness of the robot to complement human labor in routine high tunnel operations. (expand) Assessment of profitability, socioeconomic barriers for adoption and work. We have developed a draft crop budget for a harvesting robot in a tomato high tunnel. The crop budget compares manual vs robotic harvesting for several values of robotic harvesting speed and robotic lease rates.We plan to receive peer review for the crop budget and release it among farmers this year. Furthermore, our team made significant strides in engaging with urban farmer stakeholders in the last one year through participation in Dixon Springs Agricultural Center High Tunnel Production Day at Dixon Spring, IL, University of Illinois Extension Twilight Meeting Series. Simpson, IL (July 13) (attendance of 29 people), 2023 Specialty Crops Field Day. Urbana, IL (August 7) (attendance of 15 people). Furthermore the award was used to co-sponsor an 'Agricultural Robotics for Sustainable Future" workshop at the International Conference in Intelligent Robot and Systems held at Detroit, MI in October 2023.
Publications
|