Incorporating time dependent link costs in multi-state supernetworks

Multi-state supernetwork represents a promising approach to model multi-modal and multi-activity travel behaviour. A derived feature of this approach is that a point-to-point path through the supernetwork represents a specific activity-travel pattern. A limitation of current multi-state supernetworks are constructed in a static way since the elapsed time and other components of the links are pre-defined, which tends to cause inaccuracies in link costs in a dynamic context. To make this approach more powerful and practical, this study attempts to incorporate more fully the time dimension: link costs of travel with both private vehicles and public transport are calculated on-the-fly, and parking costs are made duration-dependent. Moreover, time windows at activity locations are taken into account. Thereafter, the supernetwork structure remains the same, but it can better capture time-space constraints and underlying activity-travel behaviour. A single objective bi-criterion label correcting algorithm is proposed to find the optimal activity-travel patter