The most popular formulation of a departure time choice model is based on the preferred arrival times of travellers. These preferred arrival times cannot be measured on the road, at least not in congested areas. Data collection on an individual level is necessary to collect these preferred arrival times. This is both time consuming and expensive, which makes the application of such a departure time choice model practically impossible. This paper examines an approach using reverse engineering to estimate the preferred departure times from actual departure times, current traffic conditions and an underlying choice model. This reverse engineering approach (REA) is applied on a small network and seems to perform reasonably well. Thus REA can be used to determine preferred departure time profiles. The approach is however sensitive to equilibrium conditions of the current traffic situation, the parameters of the underlying choice model and under some conditions inverting the probability matrix may lead to negative departure rates. For further research a maximum likelihood approach is proposed to recognise the stochastic nature of the process; this will also solve the problem of negative elements in the inverted probability matrix. For the covering abstract see ITRD E126595.
Abstract