Approximate Dynamic Programming: Designing an Economically Optimal Fleet of Electric Self-Driving Cars

Abstract

In this thesis, we explore a fleet of electric self-driving cars as a competitor to ride-sharing services similar to Uber or Lyft. We present a model for dispatching the cars in the fleet, and a model for pricing the fleet. Then, we explore how changing various parameters and making tweaks to how the algorithm executes affects the profits earned by the fleet. Our ultimate goal is to determine the optimal values of parameters for the fleet, such as the ideal size for the fleet, the ideal battery size of the fleet cars, and the proper recharging rate to maximize profits, and to determine an optimal algorithm for managing the fleet, also to maximize profits. We also explore how to optimize profit if we are forced to run the fleet with at least one parameter being sub-optimal, and we also investigate patterns as we hold one parameter fixed and optimize the others. Ultimately, fleets of self-driving cars present an alternative to current day ride-sharing, and this thesis investigates the profits and feasibility of such a fleet.