Simulation is being extensively used by the armed forces to train their troops as an alternative to carrying out expensive exercises in the field. The objective of using a computer for training is to provide the soldiers with the experience of working part of a team in various battle situations in conditions as much like those in the real world as possible. The simulation should provide a realistic training ground for the soldiers where it is possible to test the soldiers' skills in a variety of situations.
This project investigates the issues involved when you use the multi-agent system paradigm to model the battlefield scenario. Agents are used to represent a variety of different entities in the organisational heirarchy of the army, who must work together as a team in order to achieve the battle objectives. It is important to look at the structures and protocols that need to be in place for the effective and efficient accomplishment of the overall task. This will be based on the joint intentions model of Jennings [jennings 93] and the STEAM execution model of Tambe [tambe 98]. We are particularly interested in how the agents deal with uncertainty and failure and what mechanisms need to be used to ensure that the task is still accomplished despite the breakdown of one or several parts of the team plan.
[Jennings 93] Commitments and Conventions: the foundations of Coordination in Multi-Agent Systems, N.R. Jennings,
Knowledge Engineering Review (8) no. 3, pp223-250, 1993.
[Tambe 98] Implementing Agent Teams in Dynamic Multi-agent Environments, Applied AI (12), 1998.