Solving stochastic frequency-based assignment to transit networks with pre-trip/en-route path choice

This paper deals with the stochastic frequency-based assignment for transit systems, considering pre-trip/en-route path choice behaviour; this problem is relevant for (uncongested or congested) urban transit networks, where travelers may not completely know the status of service, say bus arrivals at stops, when they leave the origin; under mild conditions, travel strategy can be modelled by hyperpaths. Hyperpath choice behaviour can be described through random utility models thus properly modelling several unavoidable sources of uncertainty, which cannot be considered by the commonly used deterministic choice model. Effective methods suitable for large scale applications are proposed for solving stochastic assignment based on probit or gammit choice models, which properly model the effects of hyperpath overlapping, even though their application requires Montecarlo techniques; Montecarlo techniques based on Sobol numbers are compared with those based on the commonly used Mersenne Twister ones; several MSA-based algorithms for equilibrium assignment are discussed and compared with the commonly used basic implementation. Applications to a toy and a large scale network is also discussed.