Organization responsible: TNO

People involved:

• Martijn Neef (TNO, project coordinator, researcher)
• John-Jules Meyer (Utrecht University, promotor)
• Frank Dignum (Utrecht University, co-promotor)
• Bob van der Vecht (Utrecht University, PhD)

Project description:

The project research adaptive capabilities in multi-agent systems by investigating the concept of autonomy in agent decision-making. Results from this work will be used to create adaptive support systems for complex and dynamic environments.

Adaptive support systems are support systems that are capable of altering their role or function in an organisation according to changing operational demands. As artificial systems become more capable, their role in organisations will change too – from static information providers and task performers to versatile, pro-active support systems that can adapt to changing demands coming from the environment or the human organisation.

This project researches agent autonomy, and options for creating adjustable autonomy. The intention is to embed adjustable autonomy in the agent reasoning-model by defining reasoning rules that allows a software agent to self-adjust its level of autonomy. By granting agents self-adjustable autonomy, we intend to achieve dynamic coordination in groups of agents. Such dynamic coordination can be employed to creative adaptive systems.

The project should lead to both a) a better understanding of how adaptive support concepts can be developed and incorporated in future applications, and b) concrete architectural designs for agent-based adaptive support systems. To achieve these goals, this project includes both theoretical and experimental research, but uses a pragmatic, bottom-up research stance. The main research is on autonomy in agent decision making, and all other activities either stem from that topic (such as work on dynamic coordination in multi-agent systems and human – agent task delegation), or add to knowledge of the topic (such as technical facilities such as 3APL and Spyse, and experiments).

This research uses the 3APL reasoning model for agents, and adds specific extensions that enable self-adjustable autonomy. By allowing agents to self-modify their behaviour, the researchers aim to achieve new levels of adaptivity in agent-based systems.

Publications: