The Impacts of Capacity Drop on Control of Freeways: Model and Simulation Analysis

*PhD Defense*
07/26/2018 10:00–12:00
ITS Seminar Room 4080 AIR Building
Felipe Augusto de Souza
Felipe Augusto de Souza
TSE PhD Candidate

An unfortunate feature of freeway traffic flow at merge bottlenecks is the capacity drop (CD) phenomenon. It refers to a drop in the bottleneck outflow when a queue forms upstream to that bottleneck compared to the outflow observed before the formation of the queue. While its causes and exact mechanism are still open questions, this research concerns in the impacts of CD and how to mitigate them.

The distinct features of CD in a freeway corridor are assessed based on the behavior of equilibrium states in a model capable of replicating CD. The impacts are unveiled by comparing the system properties with and without the CD. The main finding is that the highest outflow occurs under uncongested equilibrium; however, it may not be reachable depending on the demands and initial conditions.

The local ramp metering control is investigated into more details. CD imposes a hysteresis on the system response with respect to the demand level. Also, we analyze the system in closed loop considering ALINEA, a well-known control algorithm. We establish the stability range with respect to parameters which is a necessary requirement for the controller to be effective. Further, we propose an extension of ALINEA to enlarge the stability range mitigating a performance loss that occurs when the on-ramp and the bottleneck are far apart.

Essential aspects of ramp metering are better captured with microscopic models; however, there were few evidences that such models can replicates CD. To that end, we propose a parameter calibration procedure that ensures the underlying model properly captures CD. The approach is tested with loop detector data from a merge bottleneck in which the CD is consistently observed. Finally, the analytical results are validated, and further practical aspects, such as detector placement and the effect of the controller sample time, are studied.

All results with different approaches point to the direction that the existence of CD imposes additional challenges on the system control. Fortunately, in most cases the effects of CD can be mitigated with a properly designed control strategy, such as the ones tested and proposed in this research.