Ad hoc Networking with Swarm Intelligence

University Of Delaware, Newark DE

Investigators

Abstract

This research project aims at applying the swarm intelligence framework to design a new class of networking protocols for mobile wireless ad hoc networks, facilitated attractive features such as scalability, autonomy, robustness, and fault-tolerance for routing, topology control and energy conservation. The basic idea is inspired by the collective activities of social insects, and utilizes small packets equivalent to ants in the real world. These packets, traveling like ants, deposit some information at nodes they visit similar to the way ants lay pheromone trails. This information, in turn, affects the behavior of other ant packets. With this form of indirect communication, the deployment of ant-like packets resembles a distributed reinforcement learning system that evolves itself to an optimal state, based on the current condition of the environment. This project will characterize the behavior and study the performance of swarm intelligence for ad hoc networking protocols based on probability, reinforcement learning, and stochastic control theories, in addition to large-scale simulation studies. To cope with network dynamics and mobility efficiently, the ad-hoc networks based on swarm intelligence are expected to learn and inherently adapt to changes in network states.