A method of determining an optimum phase plan in a big multiple network of controlled intersections, when part networks within the network have already been optimised independently of one another is discussed. The model gives a linear function which is minimised, with linear restrictions partly in continuous, partly in whole-number limited variables. This problem can be solved with a mixed-integer linear programming routine. It can be proved that the restrictions matrix has the unimodular property, so that a pivot transformation can be greatly accelerated. The model is particularly suitable for an integral approach to a big network by step-by-step optimation, in which the choice of part networks within the networks, part networks within these part networks, etc. permits the reduced network that is being optimised to be brought into alignment with the available computer facilities or desired processing in time.
Abstract