ChatCEA: a knowledge-driven intelligent service agent for controlled environment agriculture
Computers and Electronics in Agriculture · 2026-07 · 248 · 111733
DOI: 10.1016/j.compag.2026.111733 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computers and Electronics in Agriculture, 2026-07. Verfasst von Fulin Xia und 8 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 41 Literaturangaben. Gefördert von: Key Research and Development Program of Zhejiang Province, National Key Research and Development Program of China.
Abstract (Originalsprache)
AbstractControlled Environment Agriculture (CEA) faces challenges in accessing precise, actionable knowledge for efficient production management. Traditional expert systems and search engines often lack the flexibility and contextual understanding needed to address complex queries. Large language models (LLMs) exhibit strong generative and comprehension abilities, presenting an opportunity to enhance CEA practices. In this study, we introduce ChatCEA, an intelligent service agent framework that combines LLM with domain-specific knowledge through Retrieval-Augmented Generation (RAG). The framework consists of three key components: a two-source knowledge base module, a retrieval module for contextual search, and a knowledge integration module for handling unstructured data and information matching. The domain knowledge base complements the specialized CEA knowledge that is lacking in the generic LLM. We establish the first CEA-specific question-answering benchmark including 6 domain topics to evaluate performance. Results show that ChatCEA outperforms both search engine (Google) and general-purpose LLM (GPT-4). Specifically, the knowledge integration module enhances ChatCEA ability to address quantitative queries, achieving higher accuracy with an answer correctness of 0.71 (compared to 0.41 for GPT-4) and a 4.3 LLM-based evaluation (vs. 3.8 for GPT-4), while delivering responses that are 62% more concise across all CEA-related topics. An analysis of the contributions of management manuals and journal papers to the ChatCEA knowledge base highlights the significance of knowledge diversity. This study provides a knowledge-driven, domain-specific LLM framework that offers a reliable tool for CEA production management. The code for ChatCEA is available at https://github.com/IDEAS-ZJU/ChatCEA.
Förderung
- Key Research and Development Program of Zhejiang Province
- National Key Research and Development Program of China
Zitieren
APA
Xia, F., Pan, J., Zhong, R., Liu, W., Chen, T., Zhu, L., Yang, Y., Hua, G., & Lin, T. (2026). ChatCEA: a knowledge-driven intelligent service agent for controlled environment agriculture. Computers and Electronics in Agriculture, 248, 111733. https://doi.org/10.1016/j.compag.2026.111733
Vancouver
Xia F, Pan J, Zhong R, Liu W, Chen T, Zhu L, et al. ChatCEA: a knowledge-driven intelligent service agent for controlled environment agriculture. Computers and Electronics in Agriculture. 2026;248:111733. doi:10.1016/j.compag.2026.111733
BibTeX
@article{Xia2026,
author = {Xia, Fulin and Pan, Jiandong and Zhong, Renhai and Liu, Wei and Chen, Tongpeng and Zhu, Linchao and Yang, Yi and Hua, Guichao and Lin, Tao},
title = {ChatCEA: a knowledge-driven intelligent service agent for controlled environment agriculture},
journal = {Computers and Electronics in Agriculture},
year = {2026},
volume = {248},
pages = {111733},
doi = {10.1016/j.compag.2026.111733}
}
EUVIMED