Abstract:

This paper aims to examine and design operational strategies  for mixed flows of autonomous vehicles (AVs) and human-driven vehicles  (HVs). We propose a stylized model wherein utilitarian individuals  either drive HVs or take privately operated AVs that are collectively  dispatched. In the baseline mixed policy whereAVs and HVs share the same  lanes, we find that AVs under-join the traffic (queue) while HVs  over-join. We identify a “crowding-out effect" such that AVs will tend  to mitigate the over-joining HVs and reduce overall  congestion/throughput. To improve the traffic efficiency of the baseline  mixed policy, we consider a fully dedicated policy in which both AVs  and HVs are segregated to different lanes and a partially dedicated  policy in which only AVs enjoy dedicated lanes. We find that dedicated  policies outperform the mixed policy in both social welfare and the  aggregate throughput when the platooning effect is moderate or strong.  Exact conditions are derived for the selection of fully dedicated policy  and partially dedicated policy. Furthermore, we find that a carefully  designed subsidy is necessary for the dedicated policy to simultaneously  improve the social welfare and the throughput when the platooning  effect is very weak. These results shed interesting light on the policy  regulation for the emerging mixed autonomous paradigm: a dedicated  policy with proper lane design and subsidy (if necessary) will improve  both social welfare and aggregate throughput.