Our goal is to develop algorithms to deploy a fleet of vehicles for Mobility-on-Demand in large road networks governed by rules of the road.
Consider a fleet of vehicles providing a mobility-on-demand service across a city, where we consider rich demands such as "Pick me up at Vassar St @ Main St intersection, go to Prudential Mall or Faneuil Hall market, and drop me off at Boston Common within 50 minutes". In addition, the vehicles must take into account the rules of the road and traffic. These scenarios present difficult scalability challenges with respect to: 1) the space and time extend of operating in large environments over long periods of time; 2) the specification complexity associated with the rule set the vehicles must follow; and 3) the number of vehicles employed in the Mobility-on-Demand system.
Complex traffic situations may require vehicles to handle cases where the rules of the road are in conflict with each other or lead to infeasible planning problems. It is dangerous to just hand off the control to the human and expect that they can deal with the problematic situation. Thus, we require planning algorithms that provide motion plans even when not all rules of the road can be enforces (e.g., driving in the left lane to avoid a construction area), preferably in a minimum violating way.
Interaction between vehicles in the Mobility-on-Demand system, where vehicles must avoid and give way to each other, raises the question of fair coordination such that all vehicles satisfy their transportation requests as good as possible.
We develop computational frameworks that take into account the rules of the road, overcome the scalability challenges, deal with complex traffic situations, and achieve social optimality.