Continuous Planning of a Fleet of Shuttle Vans as Support for Dynamic Pricing

This paper solves the problem of estimating the number and type of required resources for pickup and delivery of passengers at some time in the future. By combining optimization and sampling methods, as well as making plans based on several statistical samples, we estimate the real values for the required resources and show how the sample values converge towards the real values. Our approach combines machine-learning based demand predictions, for the number of passengers, and a route optimization engine that assigns the passengers into shared shuttle vehicles. In order to validate our method we create a baseline data that is representative of the real values. We test our approach using this baseline data, and we obtain statistically significant results.