Organization responsible: TUD

People involved:

• Robert Babuška (promotor, project leader)
• Bart De Schutter (co-promotor)
• Ton J.J. van den Boom (researcher)
• Jelmer van Ast (PhD student)

Project description:

This project aims at developing a robust approach to achieve a desired collective behavior of (possibly unsophisticated) agents interacting in real time locally with their environment and with one another. This method will enable a novel, revolutionary approach to sensing, data fusion and control in systems with spatially distributed variables, without the use of centralized approaches or the provision of a global model. The possibility to implement the technology on current or near-future systems with wireless technology is an important aim of this project. Therefore, unlike in most research on self-organizing systems and swarm intelligence, attention will be paid to the aspects of real-time performance guarantees, performance monitoring, interaction with human users, and performance optimization under limited communication range and possibilities and restrictions on local computational power. The state-of-the-art methods and principles of self-organization and swarm intelligence will be further developed to meet the above needs of this framework. The result will be a systematic design methodology and a set of corresponding algorithms along with their software implementation, ready for the adoption and use in distributed wireless networks of agents capable of sensing, information processing and actuation.

It has been demonstrated that collective behaviours of possibly unsophisticated agents interacting locally with one another and with their environment result in the emergence of useful global functional patterns. This approach also offers intrinsically distributed algorithms that can benefit from parallel computation quite easily. However, the swarm-intelligence and self-organizing techniques currently in existence are essentially restricted to meta-heuristic optimization algorithms (e.g., ant colonies and particle swarm optimization). They can be used to find approximate solutions to combinatorial optimization problems, such as the traveling salesman problem, path planning, job-shop scheduling or packet routing in communication networks. As such, these methods have not yet addressed issues that arise in the real-time and interactive application of self-organizing multi-agent systems:

• the collective gathering and fusion of spatially distributed information,
• on-line monitoring and control of the swarm performance,
• real-time interaction with human users,
• task definition and distribution under limited communication range, bandwidth and possibly also computing power.

These are the research subjects that will be addressed within this project.

Publications: