SAM-A Stochastic Assignment Model

Many previous empirical studies have demonstrated multi-routeing: the phenomenon whereby drivers travelling between the same origin and destination select a variety of routes. Assignment methods fall into two classes depending on how they model multi-routeing. Equilibrium methods use functional relationships between link cost and link flow, whilst stochastic methods take link costs to be random variables. The two best known stochastic methods, Burrell and Dial, each have some deficiencies or disadvantages. A new method SAM (ℝ5Stochastic ℝ5Assignment ℝ5Model), is described which overcomes some of these problems, and which is an analytical counterpart of the Monte Carlo Burrell method. In applications to test networks, SAM behaves well and has acceptable properties. However, there are still some remaining difficulties. These difficulties are described and suggestions made for overcoming them.