Mechanical and Civil Engineering Seminar

With the increase in connectivity and in real-time responsiveness, travelers and vehicles are becoming "real-time optimizers" of their trips. The urban mobility challenges and breakthroughs of the next decades will be marked by our ability to optimize the aggregate performance of large-scale transportation systems while accounting for how the hundreds of thousands of "real-time optimizers" will locally interact among themselves and with the infrastructure. In this talk, we present modeling and optimization methods that address this challenge. We consider a family of optimization problems that rely on the use of stochastic urban transportation simulators.

First, we consider the problem of estimating travel demand for a large-scale urban area. The design of computationally efficient demand calibration algorithms is essential for transportation practice. We present efficient algorithms and illustrate their efficiency with case studies of Berlin and Singapore. Second, we consider the design of car-sharing services for Boston and New York City. We develop methods to estimate the spatial temporal distribution of demand for car-sharing and to optimize the distribution of vehicles across the city. The methods combine detailed car-sharing reservation data, sampling techniques and a discrete simulation-based optimization algorithm.