On scenario construction for stochastic shortest path problems in real road networks

TL;DR

This paper shows that carefully selecting scenario generation methods significantly improves the efficiency and stability of stochastic shortest path computations in complex real-world road networks, reducing the number of scenarios needed.

Contribution

It introduces a scenario generation approach that outperforms random sampling in stability and efficiency for large-scale stochastic shortest path problems in real road networks.

Findings

Unbiased scenario generation outperforms random sampling in stability.

Fewer scenarios are needed for accurate results with the proposed method.

Different OD pairs and objectives may require different scenario counts.

Abstract

Stochastic shortest path computations are often performed under very strict time constraints, so computational efficiency is critical. A major determinant for the CPU time is the number of scenarios used. We demonstrate that by carefully picking the right scenario generation method for finding scenarios, the quality of the computations can be improved substantially over random sampling for a given number of scenarios. We study a real case from a California freeway network with 438 road links and 24 5-minute time periods, implying 10,512 random speed variables, correlated in time and space, leading to a total of 55,245,816 distinct correlations. We find that (1) the scenario generation method generates unbiased scenarios and strongly outperforms random sampling in terms of stability (i.e., relative difference and variance) whichever origin-destination pair and objective function is used; (2) to achieve a certain accuracy, the number of scenarios required for scenario generation is much lower than that for random sampling, typically about 6-10 times lower for a stability level of 1\%; and (3) different origin-destination pairs and different objective functions could require different numbers of scenarios to achieve a specified stability.