Start Date
2024 12:00 AM
Abstract
Vertical farming (VF) is a type of Controlled Environment Agriculture (CEA) that promises high yield, year-round production, reduced land usage, shorter supply chains, and sustainable agriculture practices. However, its operation is a challenge to small farmers. Recent artificial intelligence (AI) advances have provided opportunities to design autonomous CEA systems. Unlike classic automation, in which each component is set to operate according to a timer, the proposed autonomous system coordinates all system components based on a set of goals (e.g., yield, taste, and energy consumption). The proposed autonomous system aims to model, monitor, and control the CEA ecosystem based on big data gathered in a proposed agriculture digital twin called Agri Generative Digital Twin (AGDT) platform in the cloud. Digital twin concepts have been applied in various manufacturing applications where data collected via sensors (i.e., in the context of IoT Internet of Things) resides in the cloud. Manufacturers can ask what-if questions to forecast system performance and simulate potential production changes without interrupting the physical production systems. Similarly, the proposed AGDT would collect growing data from crops and the environmental factors and allow farmers to ask VF operational questions. Collecting data from various VF systems and incorporating transfer learning techniques in AI will meet the substantial data demands for training AI models. The trained AI models can design optimal growing CEA recipes based on multiple objectives such as crop yields, quality, and costs. Farmers/users of the proposed autonomous AI models could optimize growth production and reduce costs to generate steady incomes, thus achieving sustainable agriculture practices.
Keywords
Vertical farming (VF), Controlled Environment Agriculture (CEA), Artificial Intelligence (AI), Machine Learning (ML), Digital Twin (DT), Autonomous System
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Recommended Citation
Chang, Shing I.; Prichard, Michael; Guo, Xiaolong; and Wang, Lizhi (2024). "Overcoming AI Model Training Barriers Towards Autonomous Controlled Environment Agriculture Systems," Urban Food Systems Symposium. https://newprairiepress.org/ufss/2024/proceedings/1
Overcoming AI Model Training Barriers Towards Autonomous Controlled Environment Agriculture Systems
Vertical farming (VF) is a type of Controlled Environment Agriculture (CEA) that promises high yield, year-round production, reduced land usage, shorter supply chains, and sustainable agriculture practices. However, its operation is a challenge to small farmers. Recent artificial intelligence (AI) advances have provided opportunities to design autonomous CEA systems. Unlike classic automation, in which each component is set to operate according to a timer, the proposed autonomous system coordinates all system components based on a set of goals (e.g., yield, taste, and energy consumption). The proposed autonomous system aims to model, monitor, and control the CEA ecosystem based on big data gathered in a proposed agriculture digital twin called Agri Generative Digital Twin (AGDT) platform in the cloud. Digital twin concepts have been applied in various manufacturing applications where data collected via sensors (i.e., in the context of IoT Internet of Things) resides in the cloud. Manufacturers can ask what-if questions to forecast system performance and simulate potential production changes without interrupting the physical production systems. Similarly, the proposed AGDT would collect growing data from crops and the environmental factors and allow farmers to ask VF operational questions. Collecting data from various VF systems and incorporating transfer learning techniques in AI will meet the substantial data demands for training AI models. The trained AI models can design optimal growing CEA recipes based on multiple objectives such as crop yields, quality, and costs. Farmers/users of the proposed autonomous AI models could optimize growth production and reduce costs to generate steady incomes, thus achieving sustainable agriculture practices.
