A trajectory-mining approach to derive travel time skim matrix in dynamic traffic assignment

The travel impedance skim matrix is one of the essential intermediate products within transportation forecasting models. It reflects interzonal travel costs and is explicitly utilized in trip distribution and mode choice models. The traditional method of obtaining skim matrices is to execute multiple iterations of all-to-one or one-to-one time dependent shortest path algorithms. However, the computational and memory usage limits can be easily reached when dealing with mega-scale networks such as those with thousands of zones. This paper proposes two new ideas of extracting the interzonal travel impedance information from the already existing vehicle trajectories. The numerical experiments highlight huge potential advantages of the proposed approaches in terms of saving both memory and CPU time.