Multiobjective Optimisation on Network Models

A specification, formulisation and implementation for multiobjective optimised route planning on a transport network is described. Transportation problem search spaces are large and often complex, involving many (independent) variables. Travellers often settle for any journey specification satisfying some weak set of conditions. The approach described provides provably optimal journey specifications, minimising on n variables, using an approach adopted by evolutionary computation. Also described is an evolutionary scheme to develop more efficient public transport networks. Public transport networks can be considered as ‘masks’ overlaying physical, static networks. We define constraints and conditions whereby new network masks may be evolved and tested for efficiency, possibly highlighting shortfalls in current network structure. The major advantage of this multi-criteria analysis is its capacity to take account of a whole gamut of differing, yet relevant criteria, even if these criteria cannot be related to monetary outcomes (particularly in the case of externalities or intangibles). This novel synergy between two rapidly emerging fields of optimisation (mutli-criteria optimisation) and algorithmics (dynamic graph algorithms) is described. Many problems in route planning on networks (such as transportation networks) involve many objectives, some easily quantified using standard measurement scales and others loosely specified. We believe that the work described here may be generalised to other general network optimisation problems in GIS (Geographic Information Systems) and Operations Research.