Distributed Topology Organization and Transmission Scheduling in Wireless Ad Hoc Networks

Thumbnail Image


umi-umd-1885.pdf (1.44 MB)
No. of downloads: 2379

Publication or External Link






An ad hoc network is a set of nodes that spontaneously form a multi-hop all-wireless infrastructure without centralized administration. We study two fundamental issues arising in this setting: topology organization and transmission scheduling.

In topology organization we consider a system where nodes need to coordinate their transmissions on a non-broadcast frequency hopping channel to discover each other. We devise a symmetric technique where two nodes use a randomized schedule to synchronize and connect in minimum time. This forms the basis for a topology construction protocol where a set of initially unsynchronized nodes are quickly grouped in multiple interconnected communication channels such that the resulting topology is connected subject to channel membership constraints imposed by the physical layer.

In the transmission scheduling problem we consider Time Division Multiple Access (TDMA)the network operates with a schedule where at each slot transmissions can be scheduled without conflicts at the intended receivers. TDMA can provide deterministic allocations but typically relies on two restrictive assumptions: network-wide slot synchronization and global knowledge of network topology and traffic requirements.

We first introduce an asynchronous TDMA communication model where slot reference for each link is provided locally by the clock of one of the node endpoints. We study the overhead introduced when nodes switch among multiple time references and propose algorithms for its minimization.

We then introduce a distributed asynchronous TDMA protocol where nodes dynamically adjust the rates their adjacent links via local slot reassignments to reach a schedule that realizes a set of optimal link rates. We introduce fairness models for both links and multi-hop sessions sharing the network and devise convergent distributed algorithms for computing the optimal rates for each model. These rates are enforced by a distributed algorithm that decides the slots reassigned during each link rate adjustment. For tree topologies we introduce an algorithm that incrementally converges to the optimal schedule in finite time; for arbitrary topologies an efficient heuristic is proposed.

Both topology organization and transmission scheduling protocols are implemented over Bluetooth, a technology enabling ad hoc networking applications. Through extensive simulations they demonstrate excellent performance in both static and dynamic scenarios.