Agent Oriented Distributed Simulation

Principal Investigator:

Starting Date: 2001-09

Duration: 3 years

Overview:

Distributed simulation has become increasingly important in recent years as a strategic technology for linking simulation components of various types at geographically different locations to create a common virtual world. Multi-Agent Systems are being used increasingly in a wide range of application areas, including business process modeling, military simulations, and Internet games. An agent can be regarded as an encapsulated computer process that is situated in an environment and which is capable of flexible, autonomous action in that environment in order to achieve its goals. An agent can receive inputs from the environment through sensors and act on the environment through effectors. It may also communicate with other agents via some form of communication language and typically has the ability to engage in cooperative problem solving. This project aims to develop autonomous agents for distributed virtual worlds, including complex life-like or game-like simulated environments. Such agents must be able to generate and monitor plans in an environment where only partial information is available and decisions must be made under time pressure. For example, in a battlefield simulation, an agent may need to make decisions as to the best route to take while uncertain about the exact position of the enemy. Agents must be proactive in being able to generate new plans in anticipation of future goals, as well as reactive, in monitoring and changing existing plans. They must also be able to pursue several goals at the same time, while responding appropriately to changes in their environment.Secondly, the project aims to support the distribution of agents and their environment across multiple processors, avoiding the bottlenecks that result from a single centralized server. To achieve efficient performance and scalability of the distributed system, a number of research issues need to be addressed including mobility of agents and dynamic load balancing, dynamic data distribution management as the interaction partners of agents change and synchronization algorithms to preserve causality. A key problem is the management of the shared environment, which needs to be structured in such a way as to allow agents to interact with the environment in a causally consistent way