Concept and Development of a Multi-Agent Digital Twin of Plant Focused on Broccoli.

The paper discusses the principles of developing a multi-agent digital twin of plants using broccoli as an example of plants. The developed model of the digital twin of plants must meet the following requirements: real-time environmental data acquisition, user feedback collection, continuous adaptation of the plant development plan for each event, individual instance for field or field part. The digital twin of plant is designed as an intelligent cyber-physical system that has a user-defined knowledge bas and a multi-agent system for planning and modeling of plant growth and development, as well as for forecasting crop parameters. For this purpose, a new method for estimate stage duration and yield is proposed, which defines a "tube" – a corridor to each of the factors corresponding plant development. The key factors have been determined during consultations with practicing agronomists but can be adjusted by users experience. This concept was originally introduced for wheat digital twin, but now is scaled and modified to simulate broccoli growth process.