Development of a software architecture for agent-based modeling of intelligent agricultural systems

This article presents the architecture of an agent-based modeling software package for intelligent agricultural systems, focused on modeling the interactions between robots, plants, and infrastructure in an apple orchard. The system integrates physical, sensor, effector, energy, and computational models into a single discrete 3D environment and supports decentralized federated learning without a centralized server. Particular attention is paid to agent autonomy, asynchronous simulation execution, and the ability to integrate with real sensors and robots. Aim. The study aims to develop the architecture of an agent-based modeling software package designed for simulating intelligent integrated information and control systems in a real, physically correct, dynamic, and partially observable environment. Research methods. The primary research method is agent-based (multi-agent) modeling, which allows simulating the interaction of autonomous agents in an uncertain and dynamic environment. Object-oriented design using UML notation is used to structure the architecture and decompose tasks. Results. A software architecture is proposed that takes into account entities such as a simulated World, Agent, Entity, Billboard, and Computer. Conclusions. The proposed platform ensures the reproducibility of experiments, scalability, and serves as a basis for testing collective behavior algorithms in heterogeneous and resource-limited agricultural environments.