Effective and economical management of herbicide-resistant weeds continues to be one of the greatest challenges for soybean production. Adapting site-specific technologies can result in more efficient herbicide applications and reduce the reliance on broadcast postemergence herbicide applications by directing herbicide applications. Field research will be conducted in 2023 and 2024 to evaluate methods for developing prescription maps for variable rate soil residual herbicide applications and document the overall benefit to weed management. In addition, an integrated system that uses both unmanned aerial systems (UAS) and unmanned ground vehicles (UGV) will be developed for autonomous weed control. The system will be evaluated to assess efficiency of various tasks such as weed identification, weed mapping, information sharing among UAS and UGV robots, crop avoidance, ground robot path planning, and weed control through mechanical/chemical/other means. Each task will first be assigned to either a UAS or UGV. Then, the individual tasks will be modeled via separate neural networks. Finally, the neural networks will be optimized with collaborative awareness using deep reinforcement learning algorithms (DRL).
Our short-term goals are to improve weed management by: 1) optimizing application methods for soil residual herbicides with emerging site-specific application technology, and 2) developing automated weed scouting and ground control robots that reduce reliance on postemergence herbicides. The benefit of site-specific application of soil residual herbicides could be realized by soybean farmers within the next three years, while the automated weed management system with UAS and UGV robots would have benefits realized long-term as the technology advances.

Project Monitoring and Evaluation Approach:
1) Develop prescription maps for variable rate application of soil residual herbicides for commercial Indiana fields with a comparison of total herbicide load.
2) Assess weed emergence density and overall weed management in standard and variable-rate soil residual herbicide applications.
3) Develop maps to document weed identity, density, and location with UAVs.
4) Use maps developed by UAV to direct weed control efforts by UGV robots.

The KPIs of the variable rate application of soil residual herbicides will be measured by the development of prescription maps for herbicides, successful variable rate application, and assessment of weed management. These indicators will be evaluated at the end of Year 1 to allow for any improvements prior to implementation of the Year 2 research in rotational crops and a potentially different suite of soil residual herbicides in subject fields. These KPIs will be evaluated again at the
conclusion of the research after Year 2.

Thorough simulations will be conducted with varying field conditions and scenarios before coming up with a final system that can be transferred to a UGV. For the overall system that can operate in fields, the most crucial performance indicators will be time taken to scout per acre and the number of weeds missed per acre by both UAS and UGV. The system should overall reduce the time taken by manual scouting and additionally require less effort for precision weed removal. In addition, The KPIs of the variable rate application of soil residual herbicides will be measured by the development of prescription maps for herbicides, successful variable rate application, and assessment of weed management. These indicators will be evaluated at the end of Year 1 to allow for any improvements prior to implementation of the Year 2 research in rotational crops and a potentially different suite of soil residual herbicides in subject fields. These KPIs will be evaluated again at the conclusion of the research after Year 2.

Variable Rate Soil Residual Herbicides:
- Prescription maps for variable-rate application of soil residual herbicides developed for target fields (March 2023).
- Successful variable rate application of soil residual herbicides in target fields (May 2023).
- Completion of evaluation/analysis of weed management, economics, and environmental loading from herbicides (September 2023).

Automated Weed Management:
Year 1 Deliverables
- Develop a simulation environment for testing out navigation capabilities of UAS and UGV - Evaluate traditional navigation algorithms for UAS and UGV navigation
- Develop DRL navigation algorithm and compare its performance over conventional algorithms
- Identify and implement other tasks such as information relaying, object avoidance, weed identification using neural networks
Year 2 Deliverables
- Develop a collaborative DRL algorithm which combines all the identified tasks
- Translate and transfer the developed algorithm from simulation to real UAS and UGV
- Test the UAS and UGV in agricultural field

The project findings will be integrated into traditional extension publications, videos, and the Purdue Weed Science website as permanent records of the information generated from this project. Site-specific weed management through variable-rate soil-applied herbicide applications and the advancement of UAS and UGV robotics for weed management are being developed in collaboration with industry partners to strengthen the path to commercial adoption and impact. The greatest impacts of this project would be modifying the traditional approach to the use of herbicides to improve stewardship, enhance field scouting for improved IPM with UAVs, and provide alternative ground tactics to control weeds that reduces our sole reliance on herbicides for weed management.

Update:

View uploaded report

The primary target audience for the information generated will be Indiana farmers and the crop production industry. Key deliverables will be identified during and upon conclusion of the project. We will employ several methods to disseminate information to soybean producers, including traditional grower contact points (field days, winter meetings, CCA training and research update events, Pest and Crop Newsletter articles, ag media, videos, and the Purdue Weed Science website). We also anticipate that industry partners involved with adapting this technology or further developing methods will be involved with communicating our findings. Site-specific and automated weed management tactics will highlight the progressive nature of Indiana soybean and corn farmers for developing and delivering practices that provide more sustainable pest management and further improve environmental stewardship.