Browsing by Subject "Maximization"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Pressure-based dispatch for shared autonomous vehicles
    (Center for Transportation Studies, University of Minnesota, 2019-08) Levin, Michael; Kang, Di
    Shared autonomous vehicle (SAV) technology is rapidly maturing, with two companies (Uber and Waymo) already testing SAV services in cities in the US. Due to the point-to-point service and lack of a driver, SAV service costs could be similar to that of personal vehicles, resulting in major mode choice changes for daily travel. A major issue for SAV operators is the SAV dispatch problem, i.e., how to optimally assign vehicles to waiting passengers. SAV dispatch is essentially a vehicle routing problem, which is NP-hard, and is complicated by fleets measured in thousands of vehicles in typical cities. Previous studies have attempted to quantify the number of passengers served per SAV using agent-based simulation studies on realistic networks, with a variety of results that highly depend on the heuristic chosen for SAV dispatch. Ideally, the optimal SAV dispatch strategy would serve as many passengers as any other policy. This project created a max-pressure dispatch policy, which was analytically proven by showing stability in the number of unserved passengers through a Lyapunov function. Essentially, the work analytically compared the serviceable demand from the max-pressure dispatch to the demand that could be served by any other dispatch policy. The max-pressure policy relied on a planning horizon; as the horizon grows to infinity, the policy becomes arbitrarily close to any sequence of SAV movements that can serve given demand rates.