An event-driven multi-agent middleware architecture and protocol design for intelligent geographically distributed battlefield training, modeling and simulation

In this paper we present a distributed event driven middleware architecture for situational awareness and intelligent decision making for command and control of geographically distributed networked battlefield agents. We tackle the important and challenging issues of distributed agents scheduling, synchronization, load balancing, and terrain database distribution/management/allocation in a distributed virtual battlefield environment. Here, battlefield agents have a limited knowledge of the global terrain environment and exchange information over a distributed network. We propose a terrain partitioning and dynamic scheduling algorithms, where scheduling of agents on different nodes depends on the geographic distribution and locations of the agent entities in the terrain. In the proposed algorithm, agents with common goals and interest are grouped together to run on the same computing node. The aim here is to reduce the communication network latency cost between different nodes and to increase efficiency and overall system computational performance. In particular, the proposed scheduling and load balancing algorithm makes use of the idea that entities that are geographically located close to and closely related to each other, e.g all entities in the same troop, will communicate more with each other. In this instance, scheduling the agents representing these entities to run on the same node will reduce the communication cost. We also propose a communication protocol between the networked distributed agents as well as describe a method to tackle the important issue of battlefield terrain division and dynamic allocation/re-allocation of resources in the distributed computing environment. Finally, experimental results are presented to show the value and potential of the proposed method.