Мультиагентна система ІНТЕЛЕКТУАЛЬНИЙ АВТОНОМНИЙ РІЙ БПЛА: SWARM-UAV
Software system for controlling an intelligent autonomous swarm of unmanned aerial vehicles based on a hybrid multi-agent architecture. The system implements a three-level management hierarchy (Swarm Governance, Swarm Executive, Agent Controllers), which combines symbolic methods of artificial intelligence (BDI agents, Behavior Trees) with neural components (Graph Neural Networks, Multi-Agent Reinforcement Learning).Includes modules of decentralized distribution of tasks through the auction mechanism, coordination of specialized groups of agents (scouts, attack aircraft, navigators), fault tolerance with limited recovery time, adaptive training of policies and communication protocols. The architecture provides six classes of formal guarantees: security, budget correctness, optimality, survivability, no hunger, and inter-level consistency.The system is based on 17 architectural contracts with mathematically proven properties.
And students and graduate students of the department take an active part in the project.