Proceedings 3rd NASA/IEEE Workshop on Formal Approaches to Agent-Based Systems (FAABS-III) (original) (raw)

The problem of maintaining a desired number of mobile agents on a network is not trivial, especially if we want a completely decentralized solution. Decentralized control makes a system more r e bust and less susceptible to partial failures. The problem is exacerbated on wireless ad hoc networks where host mobility can result in significant changes in the network size and topology. In this paper we propose an ecology-inspired approach to the management of the number of agents. The approach associates agents with living organisms and tasks with food. Agents procreate or die based on the abundance of uncompleted tasks (food). We performed a series of experiments investigating p r o p erties of such systems and analyzed their stability under various conditions. We concluded that the ecology based metaphor can be successfully applied to the management of agent populations on wireless ad hoc networks. 3 Problem Formulation 3.1 Motivation In a typical dynamic ad hoc network there is limited, variable bandwidth between hosts, and the memory and CPU on each host is constrained. Given this dynamic and resource constrained environment, it is impractical to prescribe any precomputed solution. The solution we propose for such networks is to create a system that can control the number of agents dynamically, adapting to the ever-changing environment. In order to work in the context of an agent based system, a control system should be distributed and decentralized. By distributed, we mean that the system should be able to use the underlying network to parallelize problem solving on multiple hosts. By decentralized, we mean that the system should avoid reliance on a single node, and should allow each agent to act independently. The emergent behavior resulting from the individual localized control decisions ideally will yield an optimal, or near-optimal, solution at the global level. 3.2 Approach Large ecosystems usually have several attractive qualities (such as dynamic decentralized control, self regulation, no single point of failure, robustness, and stability) that we require for our system. We propose a solution to the problem of determining the number of agents appropriate for a task at hand that is inspired by large ecosystems: 1. Each task in our system is associated with food. 2. Agents which successfully complete a task collect the associated food points. 3. Agents consume food points over time to sustain their existence. 4. Agents that exhaust their supply of food die. 5. An abundance of food can cause a new agent to spawn. I " All of these techniques promise to improve on the current research and provide a more stable decentralized ways to control the number of agents on a wireless ad hoc network. 5.2 Conclusions This paper developed an ecology-based model for managing the number of agents on ad hoc wireless networks. We have discovered that an ecosystem based model can provide decentralized distributed robust control of agents in dynamic and uncertain network environments. Our approach involves a novel exploitation of properties of ad hoc networks, enabling mobile agents to automatically adapt to changes that affect their communication and migration. The capability to dynamically adjust to the state of their network provides new possibilities for stable MAS.

Sign up for access to the world's latest research.

checkGet notified about relevant papers

checkSave papers to use in your research

checkJoin the discussion with peers

checkTrack your impact