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

Stochastic shortest path (SSP) computations are often performed under very strict time constraints, so computational efficiency is critical. A major determinant for the CPU number of scenarios used. We demonstrate that by carefully picking right scenario generation method finding scenarios, quality can be improved substantially over random sampling a given scenarios. study extensive SSP instances from freeway network and an urban road network, which involve 10,512 37,500 spatially temporally correlated speed variables, respectively. On basis experimental results total 42 origin–destination pairs 6 typical objective functions problems, we find (1) generates unbiased strongly outperforms in terms stability (i.e., relative difference variance) whichever pair function used; (2) to achieve certain accuracy, required much lower than sampling, typically about 6–10 times level 1% network; (3) different could require numbers specified stability.