Hierarchical, distributed and multi-agent control for ATM B. De Schutter Control Lab, Fac. ITS, Delft Univ. of Technology http://lcewww.et.tudelft.nl/~deschutt Centralized vs decentralized control Hierarchical control Free flight Hierarchical multi-agent control
Centralized vs decentralized control on-line real-time control modular, scalable locally optimal performance instability, conflicts global performance required information bandwidth computational complexity lack of scalability
Hierarchical control compromise between centralized & distributed local controllers higher-level coordination safety & global performance tractable & scalable
Current ATM limitations predefined corridors inefficient airspace utilization indirect routing, non-optimal altitude & speed trajectories do not exploit favorable winds results: increased flight time + fuel consumption ATM complexity current ATM architecture is mainly centralized (USA) results: heavy workload for air traffic controllers, unnecessary ground and air holding, not fault tolerant growing demand for air travel
Future concepts [Tomlin&Sastry] free flight pilots determine own routes, speed, altitude restrictions imposed in congested air space: close to airport control tower aircraft coordinate to predict and resolve conflicts: decentralized ATM benefits: reduced flight times & fuel burned, increased fault tolerance & capacity hierarchical multi-agent control
Hierarchical multi-agent control
Control design techniques decision support systems: heuristics, AI, fuzzy logic, scheduling, operations research & optimization multi-agent controllers: game theory, automated learning, adaptive control, model predictive control verification: automata, hybrid systems
References (free flight & multi-agent control) C. Tomlin, G. Pappas, J. Lygeros, D. Godbole, S. Sastry, G. Meyer, “Hybrid control in air traffic management systems,” Proc. 13th IFAC World Congress (IFAC'96), San Francisco, pp. 127-132, June-July 1996. G. Pappas, C. Tomlin, J. Lygeros, D. Godbole, S. Sastry, “A next generation architecture for air traffic management systems,” Proc. 36th IEEE Conference on Decision and Control, San Diego, pp. 2405-2410, Dec. 1997.