Automated radio network design using ant colony optimization

Abstract

Radio networks can provide reliable communication for rural intelligent transportation systems (ITS). Engineers manually design these radio networks by selecting tower locations and equipment while meeting a series of constraints such as coverage, bandwidth, maximum delay, and redundancy, all while minimizing network cost. As network size and constraints grow, the design process can quickly become overwhelming. In this thesis we model the network design problem (NDP) as a generalized Steiner tree-star (GSTS) problem. Any solution to the minimum Steiner tree (MST) problem on a constructed GSTS graph will directly identify the tower locations and equipment needed to build the network at an optimal cost. The direct MST solution can only satisfy coverage constraints. Because the MST problem is known to be NP-hard, our research applies ant colony optimization (ACO) to find near-optimal MST solutions. Using ACO also allows us to meet bandwidth, maximum delay, and redundancy constraints. We verify that our approach finds near-optimal designs by comparing it against a 2-approximation algorithm in several different scenarios.