Control of Large Scale Multi-Agent Systems: A UCAV example


Datta Godbole
Honeywell Technology Center

Distributed multi-agent systems arise in many application areas, such as Intelligent Vehicle Highway Systems, Automotive Control, Flight Control and Flight Management Systems, Network of Satellites, Power Generation and Distribution Systems, Flexible Manufacturing Systems, and Communication Networks. Most of these systems are currently designed and operated by ad-hoc divide and conquer approaches in which the problem is first decomposed into smaller manageable sub-problems and solved independently. This results in very conservative designs and under-utilization of resources. Control of such complex large-scale systems poses challenging problems of both theoretical and practical significance.

We propose a three level hierarchical architecture for control of large-scale systems. The top level of the hierarchy is concerned with resource allocation among various components/agents of the system. Lower levels of the hierarchy involve local performance optimization for each agent including inter-agent coordination and conflict resolution.

We will demonstrate our approach with an integrated trajectory planning and control of a group of unmanned air vehicles. Traditional control synthesis methods use models of plant or process to be controlled, in off-line generation of feedback controllers for specific objectives such as set point or trajectory tracking. Control of unmanned air vehicles requires increased automation much beyond the traditional control systems. Autonomous operation requires both trajectory design (planning) and trajectory tracking (control) tasks to be completely automated. We use multiple models at different resolution and time-scales to achieve seamless integration of planning and control.