Evaluation of Ramp Metering Algorithms Using Microsimulation

Abstract

The goal of the thesis was to determine the effectiveness of implementing different ramp metering strategies along I-35 in Kansas City using microsimulation analysis. Ramp metering enhances traffic conditions on the mainline by restricting the accessibility of the on-ramp traffic. Traffic data for one year (04/01/2016 to 03/31/2017) during the peak period were used to evaluate the performance of the facility before and after implementing known ramp metering strategies. The evaluation was done using the VISSIM microsimulation software. The locations of the metered junctions along I-35 were obtained from the Kansas Department of Transportation (KDOT), since KDOT installed ramp meters at these locations in 2017. Four ramp meters were located at the southbound direction and two at the northbound direction. In this thesis only the I-35 southbound movements were evaluated, as the meters at the northbound direction were placed primarily for safety purposes. The I-35 southbound corridor starts from Cambridge Dr. in the north and ends at 75th St. in the south. The ramp meters are located at the 7th St., Southwest Blvd., 18th St. Expressway, and 67th St. on-ramps. Currently, KDOT is implementing a speed-based algorithm, the details of which are unknown since the exact algorithm is proprietary. As such, for the purposes of this thesis a review of the literature was conducted to identify possible ramp metering algorithms to evaluate, and it was decided to use one localized and one system-wide ramp metering algorithm. The selected localized ramp metering algorithm is the ALINEA (Papageorgiou et al., 1991). ALINEA is an occupancy-based ramp metering algorithm that operates to maintain the occupancy in the freeway at the congestion location close to the critical occupancy that corresponds to maximum throughput. The selected system-wide ramp metering algorithm is HERO (Papamichail and Papageorgiou 2008). HERO uses ALINEA as its base algorithm, and uses a master/slave protocol. These two ramp metering algorithms, as well as the No Control scenario were evaluated considering various performance measures obtained through microsimulation. Traffic data were obtained from the KC Scout portal. The data obtained were screened for days with adverse weather conditions, traffic incidents, and bad detector data. The remaining data were used to obtain traffic demands and off-ramp relative flows to be used in VISSIM. The three control scenarios (ALINEA, HERO, and No Control) were simulated using 60 demand scenarios. These scenarios were created by averaging the weekday data in each month. Each demand scenario was run four times with different seed numbers to account for variations throughout the week, resulting in a total of 240 simulated days. The selected performance measures that were used to perform the evaluation were travel time and travel time reliability, speeds, throughput, queue lengths, and congestion duration. The entire facility travel time did not show significant improvement; however, significant travel time improvements were observed at the northern part of the facility. Congestion duration decreased after implementing the ramp metering algorithms at all metered locations except the 67th Street. Mainline spot mean speed at the metered locations also increased. Also, the throughput increased after implementing the ramp metering strategies compared to the No Control scenario. Overall, ALINEA was found to perform better than HERO; however, ALINEA had longer queues on the on-ramps, spillback percentage to the arterials and waiting times compared to HERO at all the metered locations except at 7th St. on-ramp. This is because in ALINEA a queue flush system was used when the queue length reaches a threshold, while in HERO, a queue control strategy that adapts to queue length was used.