On August 14, the launch meeting for the Inner Mongolia Autonomous Region's major science and technology project, "Key Technology Development and Demonstration Application of Agricultural Multi-Agent Spatiotemporal Large Models," was held in Hohhot.
The project is led by Hulunbuir Agricultural Reclamation Group Co., Ltd., in collaboration with seven universities and enterprises, including Inner Mongolia University, Inner Mongolia University of Technology, Inner Mongolia Agricultural University, Harbin Institute of Technology, Institute of Computing Technology, Chinese Academy of Sciences, South China Agricultural University, and Xihua University. It is headed by chief scientist Professor Luo Xiwen, an academician of the Chinese Academy of Engineering from South China Agricultural University, with Academician Lu Zhanyuan, president of the Inner Mongolia Academy of Agricultural and Animal Husbandry Sciences, serving as chair of the advisory committee.
Based on the needs of the million-acre cold-region agricultural scenario of the Hulunbuir Agricultural Reclamation area, the project aims to address challenges such as traditional agricultural data silos, low efficiency in agricultural machinery coordination, and lagging crop condition monitoring. Leveraging the research capabilities of universities and institutes, it will develop core technologies for agricultural multi-agent spatiotemporal large models. This involves collecting data covering the entire process of plowing, sowing, managing, and harvesting for eight crops, including corn, soybeans, rapeseed, wheat, and sugar beets, as well as soil, meteorological, and pest/disease data. The goal is to build high-quality datasets, spectral or image sample libraries for crop diseases and pests, remote sensing image databases, and key parameter and model libraries.
The project will develop specialized agricultural large models and an integrated intelligent agricultural machinery command platform to direct the efficient operation of newly developed unmanned intelligent agricultural machinery for power, plant protection, and harvesting. It aims to establish standardized datasets and an integrated management and control platform for a 50,000-acre demonstration area. By constructing a proprietary smart agricultural technology system for the northern cold region, the project will achieve precise, coordinated production across tens of thousands of acres, improving the digitalization and intelligence level of the group's agriculture, and producing replicable demonstration results.
The project will advance along three main directions: "effect orientation, scenario verification, and integration pathways." It will build a collaborative system of "front-end research plus end-end integration," driving deep linkage across five research topics: (1) multi-modal spatiotemporal big data representation of agricultural production and construction of field scene datasets; (2) development of agricultural-specific large models and parameter inversion technology for key crop conditions; (3) development of embodied intelligent agricultural machinery integrating lightweight models and high-precision autonomous collaborative operation technology; (4) knowledge-driven standard systems for agricultural multi-agent collaborative operations; and (5) development of an integrated intelligent management and control platform for smart agriculture with large-scale demonstration at the ten-thousand-acre level.
At the meeting, heads of each research topic reported in detail on the research positioning, content, assessment indicators, and technical roadmaps, and exchanged views on key and difficult points for project implementation, such as resource coordination, technological breakthroughs, and demonstration applications.
The convening of this launch meeting marks the official entry of the project into the full implementation phase. Currently, the project team has begun orderly work on field data collection, computing infrastructure deployment, and scenario technology optimization, steadily advancing the project's implementation.
Inner Mongolia Daily reporter: Han Xueru.