Search

Freeway ramp metering has been widely employed as an effective strategy to reduce congestion and increase the freeway operational efficiency for over two decades. Over the years, a number of isolated and coordinated metering strategies have been developed and deployed various parts of the world. Minnesota, first through the Zone metering algorithm, and later through its successor, the Stratified Zone Metering algorithm, has been among the states that extensively use freeway on-ramp metering. Based on MnDOT Regional Traffic Management Center (RTMC)’s recommendation, alternatives for developing the next generation strategy to address limitations and substantially enhance the performance of the currently deployed Stratified Ramp Metering strategy were explored. Following a different approach, the Next Generation strategy was developed by focusing on density rather than flow. This is because (as shown in earlier research) while values of occupancy near capacity are quite stable, bottleneck capacity has stochastic variations and a control strategy based on flow thresholds is likely to be inefficient. This variability in the capacity flow would mean that a control strategy based on flow thresholds would be likely to either under-load the freeway during the uncongested regime of traffic flow, or overload the system after the occurrence of the breakdown. While the former might lead to early onsets of congestion (congestion not being delayed as much as possible since the full capacity of the system is not utilized), the later might mean that the system is unable to recover from congestion efficiently (due to an over-load on the system). Critical occupancy however, and therefore density, iii is known to have stable behavior at capacity. This suggests that using a density based control approach can potentially enhance the overall performance of the system. During the first part of the study we developed a methodology to estimate densities with space and time based on data from loop detectors. The methodology is based on solving a flow conservation differential equation (using LWR theory) with intermediate (internal) freeway mainline boundaries, which is faster and more accurate than previous research using only external boundaries. Capacity drop phenomenon is inherently incorporated in the density estimation process, and the effect of the stochastic nature of capacity flow is minimized by identifying bottleneck threats and zones based on critical density values. Results compared with micro-simulation of a long freeway stretch show that this model produces reliable and accurate results. We further extended this density estimator using a two-value capacity (before and after the occurrence of a breakdown) and we integrated it in the LWR formulation. By carefully analyzing empirical data of active bottlenecks in the Twin Cities Metropolitan Area we noticed that (i) there are many cases where capacity is underutilized (4 min ramp delay constraint is misinterpreted by the algorithm) and (ii) the system once congested is unable to return to a state of flow near capacity for too long. One of the main reasons for the above inefficiencies is that capacity is considered constant during all times at all bottlenecks. This is concluded based on two empirical findings: (i) a significant capacity drop after the breakdown in many locations (varying iv 0-15%) and (ii) the total capacity of a bottleneck (sum of mainline + on ramp) is a function of the ratio of the two flows. More specifically, when ramp flows are higher the capacity is smaller (~5-10%). This happens very often in MN ramps because of the 4 minute constraint in ramp delays. Instead of a layer-based algorithm, we proceed with a dynamic zone-based algorithm. The whole freeway system is divided into zones, where the length of each zone is dynamic and is estimated in real-time. Within each zone the metering rates are chosen independently of conditions in other zones. The algorithm’s goal is to keep the car density levels at all ramps below the congestion thresholds and not to allow low speeds to occur in the mainline, by constraining the ramp delays. The ramp rates become stricter when mainline density is close to the congestion threshold, and the ramp rates increase when ramp waiting times are close to the ramp delay threshold. When it is not possible to keep both uncongested because of high on-ramp and mainline demands, the algorithm seeks to delay as long as possible the violation of both thresholds. The effectiveness of the new control strategies has been assessed by comparison with the current Stratified Zone Metering (SZM) version through microscopic simulation for the H-169 site. Under the new control strategy the total travel time on the mainline decreased by 1.5%, the ramp total travel time dropped by nearly 20%, the total system (freeway and ramp) travel time decreased by about 3% and total delays decreased by 8%. This finding suggests that in this case the new strategy is very effective since it reduces not only ramp delay, but also total system delay. The results clearly indicate that the new control strategy is very effective in keeping ramp wait times below the maximum allowed and in reducing ramp delay time. Another interesting observation made by analyzing the simulation results is that the new strategy substantially reduces ramp queues, while the overall ramp delay for the peak period was reduced by nearly 30%. The effectiveness of the newly developed control strategy is then assessed using the AIMSUN traffic micro-simulator against the currently deployed strategy. The new metering strategies is deployed on a simulated network and implemented using the AIMSUN API module. The strategy is compared against the current strategy using various measures of effectiveness and is found to succeed in delaying the onset of breakdown, accelerating system recovery after breakdown, and improving the overall freeway and ramp performances (through improved speeds and throughputs and reduced delays and stoppages). A proposal for field implementation of the new strategy and of comparison studies of performance based on ‘before’ and ‘after’ studies is suggested as a follow up for the study.