Workshop Description

Recent developments in multiagent systems (MAS) have been promising in achieving autonomous, collaborative behaviour between agents in various environments. The progress can be seen in competitions that serve as benchmarks in the respective fields (e.g. RoboCup, DARPA challenge). Still, most of the approaches, both in software simulations and real world robots, have problems if the environment is dynamic and the agents have to act in real time.

Examples are obstacle avoidance with moving obstacles or world models which are composed from egocentric views of numerous agents. Another aspect is the need for quick responses. In an environment where a number of agents build a team and both single agent decisions and team collaborative decisions have to be made methods have to be fast and precise. This workshop addresses various problems that occur with respect to these issues.

Topics of Interest

The main focus of this workshop will be methods from various areas such as world modeling, planning, learning, and communicating with agents in real-time and dynamic environments. Within this general framework we particularly aim to bring together researchers to discuss the following topics: World modeling (quantitative, qualitative), coaching (one agent gives advice to a group of agents), planning with resources (especially time), cooperation between agents (robot and/or humans), communication between agents (implicit, non-verbal, or verbal one), real-time systems software issues (often ignored but important if serious about real-time issues in robotics), scalability and robotics interfacing issues (demands a great deal of support from the initial design of the system).