An analysis by researchers at the University of Delaware and the Oak Ridge National Laboratory examined the impact of optimally controlled CAVs on the performance of a transportation network. It is broadly understood that the use of CAVs at traffic bottlenecks such as merging roadways, on-ramps, and speed reduction zones can reduce or eliminate disruptions by regulating drivers' responses. CAV technologies have been shown to make traffic safer and more efficient in such scenarios; however, there is relatively less study on their impact on the overall traffic network.
The authors of the paper offer a mathematical model of an optimal control framework for CAVs at intersections, in which vehicles are guided by a centralized coordinator. Then, by applying the model to a VISSIM network that was based on a traffic environment in Newark, Delaware, they determined the benefits and limitations of CAV implementation.
The study specifically compared a baseline human-driver case with the hypothetical optimized CAV case. Four scenarios were examined:
- Fully human-operated vehicle fleet with a long following distance and without the use of the centralized controller
- Fully CAV fleet with shorter following distance and without the use of the centralized controller
- Fully CAV fleet with a long following distance and with the use of the centralized controller
- Fully CAV fleet with shorter following distance and with the use of the centralized controller
The lengthened following distance in the third scenario was intended to provide a closer comparison to the human-operated case. The authors note that the scenarios only examine 0 percent and 100 percent market penetration by CAVs, and thus do not examine interaction between human-operated and automated vehicles.
The analysis then tested the effectiveness of the control framework under varying levels of demand, ranging from 10 percent to 200 percent of current existing demand, to understand how differing levels of traffic would affect behavior.
The study found that the introduction of control zones and the optimal control framework significantly decreased following distances for both local and highway traffic. This was primarily attributed to the ability of the control framework to coordinate vehicle movements inside control zones to eliminate possible conflicts.
The control framework also sharply reduced the number of lane changes, halving the rate at which highway vehicles merged. The rate at which local traffic changed lanes was unaffected.
Average travel speed increased for both the third and fourth scenarios. The change was most significant for local traffic, which experienced an average speed increase of approximately 40 percent upon implementation of the optimized control framework. Local traffic also experienced a corresponding decrease in total travel time.
The implementation of the control framework was found to result in zero stops across the entire network. As with the speed increase, this resulted in an effect that was most pronounced for local traffic.
The use of control zones was found to allow for significantly higher demand. The simulation found that travel time generally increased linearly with respect to the number of arrived vehicles for all scenarios. However, the scenarios that did not incorporate the optimized control framework experience an inflection point at approximately 150 percent demand, after which additional vehicles incurred an exponentially increasing travel time cost. The scenario with situationally controlled vehicles--i.e., operated by human drivers except at the coordinated control zones--experienced this inflection point much later, and the scenario that featured fully automated vehicles and the use of the optimized control framework did not have an inflection point at all. Thus, it seems that the use of the coordinator enables roadways to manage much higher volumes of traffic.