Incorporating insights from signal optimization into reservation-based intersection controls

Reservation-based intersection control for autonomous vehicles has the potential to make greater use of intersection capacity. Indeed, previous studies on the first-come-first-served (FCFS) policy (which prioritizes vehicles by order of their reservation request) have shown improvements over optimized signals. However, in certain situations, such as asymmetric intersections, FCFS easily performs worse than signals. To address this issue, we propose two new reservation policies, WEIGHTED and PHASED. WEIGHTED weights vehicle delay by signal timings, and PHASED simulates a signal but allows red phase turning movements that will not cause a collision. We test these policies on a city network and an arterial bottleneck intersection subnetwork and show that PHASED performs better than WEIGHTED in some scenarios, and vice versa. Furthermore, we show that using a combination of PHASED and WEIGHTED can perform better than using either one alone for the entire network. Results show that these policies provide effective and easily implemented alternatives to FCFS for reservations.