Multi-Agent Control

Multi-agent control studies how groups of robots, vehicles, or other autonomous systems can coordinate actions using shared goals and limited local information. Classical approaches include consensus and formation control, model predictive control, and control barrier functions for collision avoidance; learning-based methods include multi-agent reinforcement learning and hybrids that combine learned policies with model-based safety layers. Researchers evaluate these methods in settings such as robot swarms, cooperative manipulation, traffic networks, and distributed energy systems, using simulations, benchmarks, and physical experiments. A central challenge is scaling coordination across heterogeneous agents and changing networks while retaining reliable safety and stability guarantees, especially when communication is sparse or learned policies encounter unfamiliar conditions.