System control by autonomous agents

Decentralised control for complex systems-of-systems

More and more systems are integrated in systems-of-systems. In these configurations, the lack of hierarchical ownership and increasing system complexity make traditional centralised controllers less suitable.

The limitations of centralised controllers can be overcome by implementing new strategies for decentralised control. The controller’s functionality is subdivided into discrete functions for autonomous entities, called ‘agents’. Each agent controls a part of the system. The agents all pursue their particular goals by interacting with each other, and through this cooperation they provide the desired system behaviour.

System control by autonomous agents

Method details


Method properties

  • System control is compositional: configuration rather than coding
  • Due to scoping, fewer system-level assumptions interfere with control design
  • The transition from centralised to decentralised control can be gradual

Construction properties

  • System functionality is divided over autonomous agents with own goals
  • Information is distributed: exchanged on demand to reach the goals
  • Protocols and interfaces are defined for information exchange
  • Agents implement the components’ control behaviour
  • Agents communicate according to the defined protocols and interfaces

Research project

Partners

Industrial partner:

  • Vanderlande Industries

Academic partners:

  • Eindhoven University of Technology
  • Utrecht University

Contact

Business manager:

Frans Beenker

Telephone:

+31 (0)88-866 5420

Email:

frans.beenker@tno.nl