Abstract :
Assigning frequencies to cells in a cellular radio network has traditionally been carried out either by hand or, more recently, using algorithms derived from the theory of abstract graphs. This second approach is realised by modelling the problem as a simple graph colouring problem, where vertices of the graph represent the cells and edges corresponding to potentially interfering cell pairs. The algorithms used were originally developed to colour unweighted graphs. In effect this means that they utilise only `hard´ interference data between cells. In this note we describe a graph colouring algorithm that can be applied quite naturally to edge weighted graphs, and thus also to frequency assignment problems using `soft´ interference data. Prior to this we review a number of existing colouring algorithms widely used today. We also discuss issues surrounding the generation of interference data, and how to measure the quality of a given frequency plan. Preliminary test results indicate that the solutions offered by the algorithm are a significant improvement on those obtained using traditional published methods with hard data input. Test results are not included in this report