Optimal Ramp Metering Control For Freeway Weaving Segments

As the social economy and urbanization develop rapidly, the number of motor vehicles increases sharply. However, the growth in traffic demand exceeds the development in transportation infrastructure, so that traffic congestion becomesone of the most critical issues.On freeways, bottlenecks can be activated during peak periods due to intensive lane changes at weaving segments, where merging and diverging areas are in close proximity. This weaving phenomenon has a significant impact on capacity, and may result in capacity drop. Ramp metering (RM) is often deployed as an efficient active traffic management strategy to relieve bottleneck severity. Previously, some research was devoted to investigate capacity estimation models at weaving segments. However, the parameters used in those methods cannot be adopted directly in active traffic management strategy design and optimization. Furthermore, no research can be found which designed a RM control strategy considering weaving impacts.To this end, in this paper, the authors firstly calibrated a micro-simulation model with field collected data. The calibration goal was to replicate capacity of a two-sided weaving segment on the Whitemud Drive, Edmonton, Canada. Then, it presented a systematic approach to perform a sensitivity analysis of the factors that could potentially impact capacity of this weaving segment, which included traffic and geometric characteristics. It was analyzed thatcapacity of the simulated weaving segment was highly sensitive to all the parameters considered in this study. A non-linear regression model for the capacity estimation was developed based on the sensitivity analysis results. These findings could lead to implications for efficient and effective freeway operational improvements through designing optimal dynamic control methods.Consequently, this study then proposedan optimal RM control algorithm using dynamic capacity for weaving segments. The ultimate objective was to minimize the total time spent (ITS) and maximize the total travel distance (TTD) for the entire network. For the sake of convenience, the authors adoptfundamental diagrams (FD) to estimatecapacity drop in a weaving segment. Then, capacity sensitivity on the demand from on-ramp and mainline was analyzed, and a capacity estimation model was established. It was applied to estimate capacity dynamically in the macroscopic traffic flow model METANET. The modified METANET model was adopted in prediction layer of a model predictive control (MPC) framework. The optimization problem was solved bysequential quadratic programming (SQP).The proposed RM algorithm was tested with and without dynamic capacity, and the achievable benefits were compared with ALINEA control algorithm. The results from simulations showed that:(1) coordinated RM algorithms perform better than local ones when there are multiple bottlenecks along a corridor;2) overloading on bottlenecks can be mitigated or even prevented when capacity drop is considered in RM algorithms;3) based on capacity sensitivity at weaving segments, local and network performance can be optimized by using RM algorithm with real-time estimated capacity. This analysis could lead to the efficient and effective field applications for the freeway operational improvements.