Operation of Traffic Signal Systems in Oversaturated Conditions, Volume 2 – Final Report (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Executive Summary Traffic congestion continues to grow significantly in the United States and throughout the world. Agencies tasked with managing traffic control systems are continually challenged with moving traffic in congested conditions and situations where the traffic demand exceeds the capacity of the system. Under this condition of oversaturation, typical traffic control strategies do not work as efficiently as necessary, particularly since the objectives are decidedly different when mobility is restricted. The results of the Traffic Signal Operation Self-Assessment surveys (http://www.ite.org/selfassessment/) indicate the majority of agencies involved in the operation and maintenance of traffic signal systems are stretched thin and challenged to provide adequate service to drivers in their jurisdictions. In this project, Kimley-Horn and Associates, Inc. (KHA), the University of Minnesota, and Virginia Tech University performed research on the mitigation of oversaturated traffic conditions on arterials and networks with traffic signal systems. Dr. Alex Skabardonis provided advisory support. This research was divided into four components: • Development of quantitative metrics for oversaturated conditions • Identification of appropriate operational objective(s) based on the observed condition(s) • Development of a methodology for generating timing plan strategies to address specific oversaturated scenarios • Development of an online tool to relate measurement of oversaturated conditions with pre-configured mitigation strategies In addition to these four areas of emphasis, the research also resulted in a rational guide for practitioners to identify oversaturated scenarios and apply appropriate strategies using systems engineering. The research focused on traffic control plans (cycle, split, offset, etc.) that can be implemented by state-of-the-practice traffic signal systems. The research did not address methodologies or issues related to freeway operations, geometric reconfiguration, re-routing, traveler information, or other strategies that seek to influence travel demand, departure time choice, or route choice. Nor does the guide explicitly address strategies or oversaturated conditions for modes of travel other than vehicles (including transit, pedestrians, bicycles, or trains). There were seven tasks in this research project: • Literature Review • Identification of Diagnosis Methods and Definitions • Identification of Strategies and Objectives • Synthesis of Interviews with Experts Operation of traffic signal systems in oversaturated conditions Page 1

• Interim Report • Preparation of Practitioner Guidance • Application of Guidance and Strategies on Test Scenarios Summary of Task 1: Literature Review Development of strategies to handle oversaturated conditions is not a new concept; however literature that could directly be used for the purpose of this project was very limited. The research team found a variety of work in both diagnosis and strategy development. In diagnosis we focused on the review of techniques for measuring queues and the degree of saturation. Research on queue estimation is dominated by input-output modeling methods. These methods are limited to estimating queues up to the point of the input detector, but not. For arterial streets this requires installation of exit-side detection in order to measure a queue that is the length of the link. Such detector installation can be cost-prohibitive. Methods for measuring the degree of saturation can identify the saturation level up to the point of saturation, but estimates of saturation above “1.0” have not been shown to be reliable, except for those estimates used by SCOOT and SCATS which are not published. In this research, we developed a queue estimation methodology and quantitative measures of oversaturation that can be measured from advance detectors given high-resolution information on the phase timing. In the review of strategies we looked at previous research on adaptive control systems, “optimal” control formulations, and various other approaches. Features of adaptive control systems are described in the literature in a qualitative manner. Concepts can be leveraged, but specific algorithms are not typically described quantitatively. Notably the features of SCOOT and SCATS that handle oversaturated conditions were found to be if…then type rules with thresholds that change some parameters or impose additional constraints on certain decision variables in the formulation. Descriptions of these features are not accompanied by research indicating their ability to be effective in the real world or even in simulation. Optimal control formulations found in the literature all require information on traffic volumes, queue lengths, or both. Volume information is the most difficult to obtain during oversaturation using state-of-the-practice detection systems, which makes most of the optimal formulations difficult to apply directly. In this research, we extended these “optimal” formulation concepts to consider roll-over of unserved volume from one time period to the next time period. Summary of Task 2: Definitions and Diagnosis A substantial set of definitions were developed. In particular, oversaturated conditions were defined as the presence of an overflow queue on a traffic movement after the termination of the green time for that movement. Higher level definitions were then developed for approaches, phases, routes, arterials, and networks. A taxonomy was developed to describe a particular oversaturated scenario in terms of spatial extent (intersection, route, network), duration Operation of traffic signal systems in oversaturated conditions Page 2

(intermittent, persistent, pervasive), causation (demand, incidents, timings), recurrence, and symptoms (storage blocking, starvation). In Task 3, each mitigation strategy that was identified was compared to each of these taxonomy attributes to identify when and where each strategy would be appropriate to apply. This taxonomy helps to guide thinking, but it is not prescriptive in nature. For diagnosis, a methodology was developed to estimate queue length from second-by-second occupancy data from advance detectors and second-by-second data for phase timing. This methodology allows measurement of queues that grow substantially upstream of the detector location. In addition, two quantitative measures of oversaturation intensity, TOSI and SOSI, were developed. These measures (TOSI and SOSI) quantify the relationship between the length of the overflow queue at an intersection approach or movement with the available green time. SOSI measures how much green time is wasted when vehicles cannot move due to downstream blockage. TOSI measures how much green time is spent dissipating overflow queuing from the previous cycle. Oversaturated routes, arterials, intersections, networks, and so on are then defined as having TOSI and/or SOSI > 0 on the constituent approaches and movements at the same time. The characteristics of how TOSI and SOSI change over time are described in Task 2 in field tests in Minneapolis on TH55. Summary of Task 3: Development of Objectives and Strategies In Task 3, we define the three broad operational objectives that cover the regimes of potential operating conditions, (1) minimizing (user) delay, (2) maximizing throughput, and (3) managing queues. The minimize delay objective drives strategies and operational principles that assume undersaturated operation and encapsulates all objectives that might be considered “effective” (or perhaps only acceptable) during undersaturated operation. Minimization of user delay is the traditional basis for most actuated coordinated signal timing in North America. Avoiding phase failures is an “equitable” traffic management policy; one that over-emphasizes the importance of light traffic movements such as left turns and side streets. It does not minimize total delay, it rather minimizes each driver’s perception of being delayed at the signal. “Progression” is an objective that blurs the lines between “minimize delay” and “maximize throughput”. Progression is achieved in signal systems by arranging for the green times to be consecutively opened (by way of setting offsets) in a desired travel direction to allow vehicles to continue through a sequence of intersections without stopping. By carefully setting the offset values, the objective of minimizing delay (equity treatment for all users) can be satisfied at individual intersections while still meeting the objective of progression and providing the “consistent user experience” that drivers’ tend to expect on arterial roads. This objective is hindered when residual queues begin to form on the movements that the control scheme is trying to progress. Offsets that were designed (i.e. forward progression offsets) assuming that no queues were present will tend to exacerbate the situation. Operation of traffic signal systems in oversaturated conditions Page 3

Minimizing delay is not appropriate when the situation is oversaturated since it is no longer possible to avoid phase failures. Thus, maximizing the number of vehicles actually served by the intersection, with respect to the vehicles presented to the intersection (the load), is a more appropriate objective. This keeps as much of the system operational as possible, perhaps delaying movements or phases where the total traffic demand is quite low. From an equity perspective, strategies that maximize throughput might be considered to “punish” light movements to benefit the greater good. This is done by moving much heavier phases for longer amounts of time and more frequently than would be expected by the typical cycle-failure minimizing actuated control approach. Throughput maximization strategies goals either increase input, increase output, or both. At some point, however, no further revision to the signal timing will increase maximum throughput, and queues will continue to grow until demand diminishes. When growing residual queues can no longer be relieved by maximizing throughput, then the practitioner's only choice becomes arranging the operation of signals within a network to prevent the queues from increasing the problem. This objective is denoted “queue management”. Since so little is known or published about methodologies for mitigating oversaturated conditions), a research methodology was developed to compare the performance of various mitigation strategies optimized for the three objectives. In oversaturated traffic scenarios, the methodology also explicitly considers the identification of the critical routes through a network of intersections. Mathematical approaches similar to O-D estimation were explored in this project and are considered experimental. The multi-objective strategy development and evaluation methodology built on previous work by Akcelik, Abu-Lebdeh, Lieberman, and Rathi in generating the principles by which the green times, offsets, and objective functions of the methodology were developed. These principles were extended by the research team in the methodology to address the critical routes within the network. However, there is still significant additional work needed to fully develop a comprehensive closed-form, analytical procedure for which a set of common pattern parameters of traditional traffic controllers (cycle, split, offset, sequence, time of day schedule, etc.) could be generated. This evaluation methodology compares all possible mitigation strategies against all possible realizations of critical route flows for both delay and throughput measures. Using Pareto analysis, non-dominated strategies are identified for each time period of each test case. Typically it was found that no specific mitigation is optimal for both minimizing delay and maximizing throughput. This research process is depicted in Figure 1. This methodology was applied to two test cases in Task 7. In the first test, one timing plan was applied to the oversaturated scenario through the entire simulation. In the second test, the optimization methodology was extended to consider the regimes of operation (loading, processing, and recovery). Three timing plans were applied and evaluated in the test case. Operation of traffic signal systems in oversaturated conditions Page 4

Figure 1. Research methodology for development and evaluation of mitigation strategies In Task 3, we also qualitatively described a wide variety of potential mitigation strategies and allocated them against to the taxonomy developed in Task 2. The strategies included phase reservice, negative offsets, cycle time adjustment, left-turn type, phase sequence, dynamic lane allocation, green flush, and phase truncation. Various combinations of these strategies were tested on the two test cases in Task 7. Summary of Expert Practitioner Interviews In Task 4, we interviewed a number of expert practitioners. All expert practitioners viewed the management of oversaturated conditions as one of the most important issues they face from day to day. They frequently relied on personal experience and trial-and-error application than running models or performing extensive analytical analysis. The focus of most, if not all, of the expert practitioners was first and foremost to make simple changes to splits or phase sequences to minimize delay before applying more complex approaches. Many of the expert practitioners stated their approach to implement strategies (where possible) was to prevent oversaturation from occurring rather than reacting to the issues after the fact. Many of the expert practitioners stated the use of lower cycle times and reducing the number of phases at oversaturated locations to improve performance. Following is a list of the strategies for individual intersections cited by the expert practitioners: Constraints:  Control  Network Conduct Pareto analysis VISSIM simulation Determine feasible critical route movements Optimal strategies and objective functions Retrieve performance measures and system states Identify scenarios (system states) Identify potential control strategies Obtain and analyze detector log data Objective functions:  Travel time  Throughput Operation of traffic signal systems in oversaturated conditions Page 5

• Re-allocate split time to the oversaturated phase. • Decrease the cycle time when more than one phase is oversaturated. More opportunities to service the queue because there is a reduced efficiency of saturation flow rate in long queues. • Increase the cycle time at the oversaturated intersection and double-cycle other adjacent intersections. • Run closely spaced intersections on one controller, like an interchange, to avoid storage of vehicles on short links. One expert practitioner mentioned attempting to run three intersections on one controller. Another expert offered a rule-of-thumb that if two intersections were less than seven seconds travel time apart, they should be run on one controller. • Use lead-lag left turns to provide more green time for oversaturated through movements. • Use phase reservice. • Use adaptive control methods. This was primarily mentioned as an approach to delay the onset of congestion and reduce the total amount of time that an intersection is oversaturated by managing conditions beforehand. • Use queue detection loops to increase the max time (select an alternative MAX2 value via logic) for an oversaturated phase. • Use queue detection to decrease phase split time (select an alternative MAX2 value via logic) when downstream link is congested. • Use split phasing based on volume detection. • Run intersection free. • Do not run intersection free (this contradicts the bullet above) as this tends to exacerbate the problems for adjacent intersections. • Change the barrier structure for certain plans by time of day (omit phases, combine movements, remove protected phases when not needed). • Run fixed time, short cycles during construction. • Run fixed time at interchanges – early return to green at other locations can actually harm progression. • Use lane control signals or signs to provide flexible capacity increases by TOD / plan. • Time detector extension times appropriately to provide swift operation. • Allow pedestrian times to exceed cycle time for infrequent pedestrian operation on side streets. Strategies for managing oversaturation on arterials and grids were less commonly reported by the expert practitioners than strategies for individual intersections. Most expert practitioners stated they applied both negative and simultaneous offsets for queue management on arterials. Double cycling and combinations of harmonics (3:2, 3:1, etc.) were reported as methods they previously used for handling atypical demand at one or more intersections. Operation of traffic signal systems in oversaturated conditions Page 6

At the network level, most expert practitioners reported metering as the predominant approach for alleviating gridlock conditions in critical areas. Identifying metering locations was suggested to be done on a case-by-case basis. Some expert practitioners did not have CBD experience to draw from, but in general the least amount of time during the interviews was spent discussing oversaturation issues related to grids. Many of the expert practitioners mentioned having tried traffic-responsive methods in the past with mixed results, particularly due to waffling or difficulty in setting up the parameters so that the actions taken by a traffic-responsive system were reasonable. Few experts reported having access to performance measures from central systems that were adequate for diagnosing problems with oversaturation. All agreed that, in general, more diagnostics from both the local controller and the central system would be helpful to improve operations. One expert practitioner described a smart archiving capability that would allow queries to be processed such as “provide the top 10 heaviest left turns in the system over the last six months” or “tell me which non-coordinated phases maxed out most often during peak period”. All of the expert practitioners agreed that reliable detection systems are critical for adequate operation. Most indicated that in their jurisdictions a fully-actuated detection scheme was most common. Several experts had critical remarks about the limitations of video detection systems and indicated that a hybrid detector (inductive + radar, video + radar, infra-red, etc.) is probably the most likely candidate to overcome the limitations of a particular detection technology. In general, the expert practitioner interviews confirmed there are a number of strategies that have been and can be applied to the management of oversaturated conditions. Most techniques were described for individual intersections. Expert practitioners tend to be more interested in solving problems than documenting effectiveness, so quantitative benefits of applying a particular strategy were not described by any expert practitioner. This lack of evidence of the effectiveness of any approach is exacerbated by the fact that automated data collection becomes problematic during congested conditions. It was clear that the experts provided an adequate solution for a given situation and then moved on to other pressing field problems instead of spending additional effort to document or measure the degree of effectiveness. Summary of Practitioner Guidance In Task 6, we developed a Practitioners Guide for applying mitigation strategies to specific oversaturated condition problems. This guide is provided as a supplement to this final report. The guide is a stand-alone document which includes some of the material from the final report, but includes original material as well. The intended audience for the guide is a practicing traffic engineer with responsibility for designing and implementing traffic signal system timing, phasing, sequencing, and scheduling. The guidance follows a systems engineering approach to problem Operation of traffic signal systems in oversaturated conditions Page 7

resolution, starting with problem characterization as illustrated in Figure 2. The goal of the initial steps of this approach is to answer several basic questions: • How many intersections and travel directions are affected? (Spatial extent) • How long does the oversaturated condition last? How does it evolve over time? How does it dissipate during recovery? (Temporal extent) • How frequently does the oversaturated condition occur? (Recurrence) • What is the cause or causes of this oversaturated condition? (Causes and Symptoms) Subsequent steps recommend that the practitioner identify the objectives and approximate regimes of operation such as the duration of loading, processing, and recovery regimes. The next step in the process is to match appropriate mitigation strategies with the size and extent of the scenario and the objective(s) that are intended to be met. The guide provides mitigation strategies and examples of effectiveness and rules of thumb for application for some strategies. Figure 2. Process of identifying and addressing oversaturated conditions The guide also covers the generic process of deploying the mitigation and evaluating the effectiveness. Step 1: Become Aware of Oversaturation Condition Step 2: Spatial extent Temporal extent Step 3: Identify Objectives Improve Throughput Manage Queues Step 4: Identify applicable mitigating strategies Identify equipment needs Step 5: Implement Strategies Step 6: Observe and measure Performance Step 7: Document Results Operation of traffic signal systems in oversaturated conditions Page 8

Summary of Test Applications In Task 7, we applied methodologies developed in this project to six test networks. Two of the tests were used in the development and testing of the methodology for developing mitigation strategies and testing those strategies using a multi-objective Pareto analysis. In the first test case we considered application of a single signal timing strategy for an entire oversaturated scenario. In the second test case, we explicitly considered the three regimes of the scenario in applying a sequence of three signal timing plans during the three operational regimes. Two other networks were used in development and testing of strategies directly related to TOSI and SOSI. TH55 was used to prove and refine the concepts of TOSI and SOSI, and test the forward-backward procedure in a relatively simple situation. The Pasadena, CA downtown network was used for developing and testing the forward-backward procedure in a stressed and complicated routing scenario. Finally, two other test cases were used to test a variety of mitigation strategies using engineering judgment and applying the guidance methodology developed in Task 6. The Windsor, ON network was also used to demonstrate the application of the if…then online mitigation strategy selection tool. All of the test applications were simulated using Vissim with either the RBC or the Virtual D4 traffic controller. While route proportions and demand flows were changed over time, no dynamic traffic assignment was used (i.e. vehicles in the simulation did not react to the congestion conditions to change their route, change their destination, or forgo travel). The characteristics of these test cases are summarized in Table 1. Operation of traffic signal systems in oversaturated conditions Page 9

Table 1: Summary of attributes of test cases Test Case Config Number of Ints Spacing (ft) speed (mph) Typical phasing Test duration Causation Types of symptoms Recurrence Critical locations Types of mitigation Components tested Reston Parkway; Northern VA Arterial with freeway interchange 14 500 to 3300 40 4, 6, 8 3 hours Demand All Recurrent 2 Cycle, splits, offsets, phase reservice, gating Timing plan development framework Post Oak area of Houston, TX Network 16 400 to 1800 30-40 2, 4, 6, 8 3 hours Demand All Recurrent 8 Cycle, splits, offsets, phase reservice, gating Timing plan development framework TH55; Minneapolis , MN Arterial 5 500 to 2600 55 4 1 hour Preemption Spillback, overflow queuing Non-recurre nt 2 Green extension, green truncation Calculation of TOSI, SOSI, and queue length Downtown grid; Pasadena, CA Grid 22 400 to 1000 25-35 2, 4, 6, 8 2 hours Light-rail; demand Spillback, overflow queuing Recurrent 5 Splits, offsets Calculation of TOSI, SOSI, and queue length Border Tunnel Entrance; Windsor, ON Small Network 9 400 to 800 25-35 2, 4, 6 45 min, 1 hour, 2 hours Incident All Non-recurre nt 1 Many Online feedback tool; TOSI/SOSI Surprise, AZ Arterial 6 2600 45 8, 6 1.5 hours, 3 hours Planned event All Both 1 Many Application of guidance Operation of traffic signal systems in oversaturated conditions Page 10

Test Cases for the Multi-Objective Pareto Analysis Two scenarios were used to develop and test the strategy development methodology. The first test case applied a single mitigation timing plan to the entire oversaturated scenario. The second test case applied a sequence of three timing plans during the scenario to address the loading, processing, and recovery regimes. The first test case analyzed an oversaturated scenario on Reston Parkway in Herndon, VA. This scenario is an arterial that intersects with the heavily traveled Dulles Toll Road. Combinations of cycle, splits, and offsets designed for operation in oversaturated conditions were tested. The timing plans were combined with either upstream metering on the critical route or with phase reservice for the northbound left turn at the critical interchange. In both cases it was found, in general, that short cycle lengths (e.g. 100s) with close to simultaneous offsets would minimize total system delay. Medium-length cycle times (e.g. 140s) were found to maximize throughput. In general, strategies that were optimized to maximize throughput and combined with upstream metering decreased total delay by 20% and increased total throughput by 15%. Strategies that were optimized to minimize total delay and combined with upstream metering could reduce delay by up to 40%, but increased throughput by only 7%. Strategies that focused on maximizing throughput and combined with phase reservice decreased total delay by 27% and increased total throughput by 22%. Strategies that focused on minimizing total delay with phase reservice could reduce delay by up to 63%, but increased throughput by only 5%. In the second test case in the Post Oak area of Houston, TX, the three regimes of operation were explicitly considered in the timing plan development process. Two combinations of critical routes were evaluated. Timing plans were then designed with consideration of these critical movements and considering a sequence of timing plan changes at the beginning of the processing regime and then at the beginning of the recovery regime. The timing plan strategies considered green flaring, phase reservice, negative and simultaneous offsets, and harmonic (2:1 and 3:1) cycling at many of the intersections in the network. In both strategies, the cycle time of the mitigations was reduced during the processing regime (from 150s to 100s or 90s) and then slightly increased during the recovery regime (from 150s to 160s). In addition, approximately half of the intersections in the network were double-cycled during the processing regime (80s or 75s cycle time) in order to manage the growth and interaction of long queues on the short network links in the interior of the network. Both strategies provided modest 5-10% improvements over the baseline strategy for total delay. The (maximize throughputmanage queuesmaximize throughput) strategy produced improved throughput on approximately 2/3 of the intersections and decreased throughput on the remaining intersections when compared to the baseline. The detriments to throughput at the intersections with reduced performance were less significant than the detriments produced by the (minimize delaymanage queuesmaximize throughput) strategy. Those locations with throughput Operation of traffic signal systems in oversaturated conditions Page 11

reductions were typically at the locations where SOSI was non-zero (portions of the green time were wasted because no vehicles could move). In addition, significant throughput improvements were found at the intersections that were double-cycled. Test Cases for Direct Application of TOSI and SOSI to Re-allocate Green Time In the test applications for direct application of the TOSI and SOSI measures, we found significant improvements were possible. In the TH55 in Minneapolis application, the TOSI and SOSI values were used to identify offset and green time re-allocation recommendations to improve the throughput performance. In the test case using the Pasadena downtown network, two scenarios were tested. The first was a single oversaturated route in one direction on an arterial. The FBP was developed and applied using the average TOSI and SOSI values measured along the route in the “do nothing” condition. The resulting adjustments to the green splits along the oversaturated route resulted in a 30% improvement to the throughput along the route. Notably because of the downstream blocking conditions along portions of the route, some of the green time splits were actually reduced, but the throughput performance along the route was still increased. In the second test, two intersecting oversaturated routes were analyzed (one southbound and one westbound). The same FBP was applied to calculate the green time adjustments along both routes. In this test, the average throughput for southbound showed no appreciable improvement but the westbound route was improved by 10%. Test Cases for the Application of the Practitioner Guidance and Online Evaluation of Mitigations Two additional test applications focused on following the process defined in the practitioner guidance. The first test case evaluated mitigation strategies for handling oversaturated conditions on two critical routes competing for access to a single capacity-limited destination. In this test, a non-recurrent incident is generated at the entrance to a border-crossing tunnel. Six different mitigation strategies including metering, dynamic lane assignment, phase omits, and green time re-allocation were tested to see if more equitable allocation of green time could be provided for the two critical routes with minor effect on the non-critical routes. This test was also envisioned to explore improvement of performance to non-critical routes that were previously blocked by vehicles on the critical routes using the standard operations. All of the mitigation strategies applied to this test case showed that more efficient use of available space could be achieved by applying metering and offset strategies to reduce the occurrence of SOSI > 0 along the critical route. The performance results for total travel time, delay, and throughput were largely inconclusive as most link-by-link performance was either not significantly different than the baseline, or performance improvements on some links were offset by performance detriments on other links. The final test case focused on a heavily traveled arterial with pre-planned special event traffic overlaid on P.M. peak traffic that is already near oversaturated conditions. Eight different Operation of traffic signal systems in oversaturated conditions Page 12

mitigation strategies were formulated and tested on this scenario. The mitigations included combinations of cycle time adjustment, green time re-allocation, negative and simultaneous offsets, dynamic lane allocation, and double cycling. In general, the results for this scenario indicated that all of the mitigations outperformed the baseline operation for both total travel time, throughput and delay measures. Most of the mitigation strategies improved the system recovery time by more than 20 minutes. Conclusions and Directions for Further Research Some key take-away findings were identified during the project. The first key finding was that identifying the critical routes through a network of intersections is the first a critical first step in identifying appropriate mitigations. The methodology we developed for designing timing plans that explicitly consider critical routes showed promise that alternative formulation for the optimization process can result in significant gains in total system performance. In addition, this methodology is one of the first we know of that can optimize both the timing plan parameters of individual timing plans and the sequence and duration of application of those plans during a scenario. There is still much effort necessary to bring this complicated and experimental methodology closer to being able to be applied by a typical practitioner much like they might run a tool like Synchro or Transyt. The second key finding in this project was the development of the TOSI and SOSI quantitative measures and the development of the heuristic Forward-Backward Process to directly compute re-allocation of green times from those measurements. This process was developed and tested in an “offline” manner in this project and shown to make improvements to an oversaturated scenario. These experiments point towards online application of the FBP to continually re-compute the green time allocation in an adaptive manner. The final key finding of the project was that it is important to consider operating the system differently during the three regimes of operation during oversaturated conditions: • Loading • Processing • Recovery During the loading regime, systems are best operated by continuing to minimize total delay or maximizing total throughput. When TOSI and SOSI become significantly > 0, strategies which manage queues on the critical routes are most effective. These strategies minimize the degree of saturation on critical routes in order to minimize SOSI and TOSI effects. Minimizing SOSI is the most important goal during the processing regime. Finally, in the recovery regime, strategies which maximize throughput are much more effective in clearing the overflow queues that were generating during the processing phase than other strategies. Operation of traffic signal systems in oversaturated conditions Page 13

Directions for Future Research It is often said that good research generates more questions than provides answers. This project generated a number of significant directions for further research. While the goal of this project was initially to develop a guide for practitioners, there were simply too many unknowns to distill a limited amount of benefits information into a guide or generate a prescriptive unified theory of operations. As such, the first future research need is to evaluate additional test cases that will illustrate the performance of certain mitigations in specific situations. In order to fully develop a comprehensive guide, at a minimum, an example application of each potential technique is needed. The second major need is to test and evaluate mitigation strategies in the real world. All of the tests that were performed during this project were done with simulation tools. It is well known that simulations have challenges in representation of real-world behaviors during oversaturation. Field testing and application of mitigations in real-world sites (among those that were tested in the simulation studies during this project) would certainly be a valuable research activity to follow this effort. The practitioner guidance could be greatly improved by development of additional “rules of thumb” and more “cook book” type design principles for mitigations. This could not be achieved during this research due to the shortage of knowledge in this area. Furthermore, the combination of mitigations into comprehensive strategies is still more art than science and the methodology developed in this project is too cumbersome and complicated for use by a typical practitioner. Development of an offline analysis tool that can develop mitigation strategies in general network structures will be a valuable future research topic. Additional research is needed on the role of thresholds, persistence time, and recovery time in the measurement of TOSI and SOSI for selection of mitigations in an online manner. In this research, we selected what seemed to be common-sense values for these parameters but did not apply any sensitivity analysis on the values of these inputs. Additional scenarios also need to be constructed and tested with the online tool for more comprehensive evaluation of the effectiveness of such a tool. To truly get such methods into real practice, system operators will have to “spec” and procure such features in upgrades or new installations of ATMS. Finally, while it was found that real benefits can be achieved through the application of fixed-parameter timing plans, it was clear that online adaptive feedback control methods would improve the operation of oversaturated systems. In particular, it appears that offsets and green time on oversaturated critical routes need to be adjusted almost every cycle to mitigate TOSI and SOSI > 0. Negative offsets can be designed for a particular value of TOSI, but if the demand rate remains constant and the green time is left constant, the queue length will continue to grow until the link is filled. Development of adaptive algorithms and logic that directly consider oversaturated conditions in actuated-coordinated systems would be of benefit to the industry. Operation of traffic signal systems in oversaturated conditions Page 14

The FBP could be extended to a rolling-horizon formulation to take another step in that direction. Much additional research would be necessary to extend the basic heuristic of the FBP to consider phase sequencing, cycle time, protected/permitted lefts, and so on. Operation of traffic signal systems in oversaturated conditions Page 15