Survivability Simulator for Multi-agent Adaptive Coordination

Keywords: Multi-agent systems, discrete simulation, simulator The growth of a distributed processing system can increase both the number and likelihood of attacks it may be subject to over its lifetime 14, 7]. This fact, in addition to the complexity inherent in such an environment, makes the survivability of large heterogeneous systems one of the most challenging research areas currently being investigatedd1]. Our goal is to create a distributed simulation system to test various coordination mechanisms allowing the elements of the system to detect, react and adapt in the face of adverse working conditions. Our assumption is that the system is composed of a group of autonomous agents. Each agent has its own local view of the world and its own goals, but is capable of coordinating these goals with respect to remote agents. To simulate these complex systems, an environment is needed which permits the simulation of an agent's method execution. To this end, we are developing a distributed event-based simulator capable of simulating the eeects directed attacks or a capricious environment have on agent method execution and recovery. 1 Challenge With the advent of open computing environments, widely networked information resources, and an equally widely networked client base, applications are evolving from highly centralized structures to distributed decentralized structures. Distribution is required to cope with the complexity of the tasks, to distribute the load, and to utilize the necessarily distributed resources or expertise used by these applications. In this setting, application interactions are also evolving and often a distributed application is actually a collection of separate applications, each possibly having other tasks outside of the given application and possibly under the management or ownership of diierent controlling entities. With increasing frequency, an agent computing paradigm is used to control these application components where each contributing application is an agent, having its own goals, func-tionality, and perhaps special computing resources. The agents interact via higher-level interfaces, exchanging high-level information such as goals and partial plans, rather than simply invoking each other's low-level methods via remote-function-call style interfaces. The agent-style control methodology lends itself to this integration task because it enables designers to explicitly represent agent goals, where agents may have multiple goals and goals for diierent client applications, and to reason about the beneets or relative importance of these goals from a self-interested perspective 8, 9, 12], a wholly cooperative perspective 6, 5], or a market-driven economic perspective 13]. While …