SIGNAL TIMING UNDER SATURATED CONDITIONS
Identifying strategies for mitigating the effects of saturated signal approaches requires an understanding of the objective of operational decisions. These objectives vary with conditions, though this variation is rarely articulated by practitioners or used in recommendations to practitioners. They are, however, implicit in the decisions made by practitioners.
Over the history of the practice, methods that have received wide use already distinguish between a range of objectives in signal control. For example, a traditional approach for calculating cycle length at isolated intersections, first taught at the Yale Bureau of Highway Traffic, determines the expected arrival platoon at a signal using the Poisson distribution. This method assumes independence between arriving cars and thus is only valid at low arrival flows. It identifies the arrivals during a cycle that will only be exceeded 5% of the time, and then provides green time sufficient for those arrivals. The objective of this function is to serve 95% of the arriving platoons, which stated another way is to minimize cycle failures. Cycle failures occur when the waiting queue is not fully served by the next green interval.
At some point, minimizing cycle failures becomes an unattainable objective. The Poisson method will drive up the cycle length, with the result that the longer red times will create longer queues to serve, which drives up the cycle even more. In the Poisson method, one iterates to find a cycle at which all the 95%-ile arrivals can be served, but those iterations fail when traffic demand reaches moderate amounts. And at those demand levels, the assumptions underlying the Poisson method are no longer valid, in that the cars are no longer behaving independently and the Poisson distribution no longer describes their arrivals.
During such conditions, most practitioners switch to a method that seeks to achieve smooth flow consistent with driver expectations. This objective may be explicitly represented by minimizing delay, stops, or some combination of the two. Delay and stop minimization is a different objective than minimizing cycle failures, because at some point serving a long departing queue causes more delay on the competing movements than it saves on the approach being served. Most traffic signal timing tools currently available seek to minimize some combination of stops and delay.
Some tools use degree of saturation on the approach as a surrogate for stops and delay. For example, the Highway Capacity Manual is based on the fundamental relationship of G/C*S, where G/C is the ratio of green time to cycle length and S is the saturation flow. This approach seeks to define the capacity of the approach, and signal timing based on this approach attempts to balance the saturation of each movement. The objective is therefore to minimize the difference in the degree of saturation for the various approaches at an intersection, and those variations are a negative performance measure within that approach.
PASSER II has as its main optimization objective signal progression along an arterial street (more on that below), but it seeks to balance the degree of saturation at each intersection to optimize splits. Both are closely related to Critical Lane Analysis, which first appeared in TRB Circular 212, Interim Materials on Highway Capacity. This approach explicitly does not minimize delay. For example, if an intersection with four approaches is served by a two-phase signal, and three of the approaches each serve 1000 cars while the fourth only serves 100 cars, balancing the degree of saturation will result in equal green times for the two intersecting roadways, assuming no left turns. But when the street carrying 1000 cars in one direction and 100 cars in the other direction is green, only 1100 cars are being served, while 2000 cars are being delayed. Minimizing delay would provide close to two-thirds of the cycle to the street with two 1000-car/hour approaches, and one third to the street carrying only 1100 cars. This approach might, however, cause congestion on the 1000-car/hour approach that is opposing the light movement.
Within coordinated systems, optimization approaches consistent with achieving smooth flow optimize to minimize delay and stops in the network or to optimize a specific smooth-flow objective, such as progression. Progression is the ability of cars to pass through a series of coordinated signals without being stopped by a red signal. All such opportunities form a band that is the portion of the cycle during which vehicles traveling at the desired speed can achieve progression through the intersections in question. As mentioned previously, PASSER II explicitly maximizes bandwidth when determining the coordination between intersections and the phase sequence at each intersection. The splits at each intersection are determined to provide balanced saturation, however.
Optimizations that seek to minimize delay and stops, maximize progression, or balance the degree of saturation, fail when an approach becomes congested. The reason is that neither delay nor saturation can be evaluated during congestion, and progression cannot be achieved through a standing queue. Saturation is limited by the capacity of the signal, and when the capacity is exceeded, saturation no longer characterizes demand. Delay equations are indefinable in congested conditions because the delay increases as long as the demand exceeds capacity. Delay equations require a stable relationship between vehicles arriving to be served and those departing, having been served. When the system is storing more and more cars because of a capacity constraint, delay will continue to increase for the subsequently arriving vehicles.
Thus, an important question is determining at what point a practitioner should change objectives from attempting to achieve smooth flow to strategies defined specifically for congested conditions, when smooth flow is impossible.
It should be noted that the boundary between shifting objectives is not fixed, and the operation can cross the boundary as a result of what practitioners do. Thus, one critical strategy apparent in the case studies is to avoid inducing congestion when smooth flow is still attainable.
This chapter discusses a series of case studies and a series of interviews with practitioners in the context of these objectives. These case studies will be evaluated on the basis of which objective was being attempted, and whether the results of those attempts were successful. Several of the case studies show the progression from smooth-flow strategies that failed to congestion strategies.
In addition to identifying strategies by their objectives, the cases studies and interviews in this chapter will be used to catalog a series of action points to reveal when a change in strategies might be critical. These action points are closely related to the definition of saturated conditions.
NCHRP 3-38(4) attempted to define saturated conditions as follows:
Figure 2-1. A repeat of Figure G-1. Figure II-1 is a graph that shows cars with respect to time. Arriving flow is shown as a curve, with the slope of the curve indicating flow. Departing flows are shown as zero during the redn intervals and as saturation flow during the green intervals. Where the departure flow catches up to the arriving flow, the departure flow follows the arriving flow. The area between these curves characterizes intersection delay.
These definitions served the purpose of that research but are not necessarily clearly related to actions that practitioners would take. A useful feature of these definitions is that it distinguishes between occasional cycle failures that will be present even in moderate traffic conditions and the sort of building residual queues that compel action by the practitioner. One problem is that the order in which they are presented do not provide a clear hierarchy of severity, nor do they clearly suggest to the practitioner to change objectives. A review of the NCHRP 3-38(4) work suggests that these categories were define mostly to link to specific traffic signal system control strategies, not all of which are currently relevant.
Also, the conditions described in NCHRP 3-38(4) are closely related to coordinated signal timing, while the purpose of this research is to explore intersection-based strategies.
Thus, these levels can be simplified, providing a hierarchy of conditions keyed to action levels:
These approaches can be characterized graphically, as shown in Figure 2-2.
Figure 2-2. Figure showing that light and moderate traffic rarely provides successive cycles where demand exceeds green capacity, but that in heavy traffic that is still less than routinely calculated capacity, the frequency and probability of success cycles failures grows significantly. When demand exceeds capacity, the probability of arrival flow exceeding green capacity is 100%. The diagram is illustrative only and is not intended to be quantitative.
Figure 2-2 is not intended to represent actual statistics, but rather to illustrate that when average demand equals approach capacity, the percent of cycles with residual queuing will be 100%. In fact, it is likely that residual queuing is a certainty even when average demand is somewhat less than capacity. When average demand exceeds capacity, a growing residual queue is inevitable.
In addition to describing the conditions at which certain actions and objectives are compelled, the case studies and interviews reveal those actions. Of particular importance is determining whether the actions taken are intended to be curative or palliative. In the medical world, doctors treat patients with the objective of achieving a cure. When a cure is not possible, then they treat terminally ill patients with the objective of minimizing pain and suffering.
The initial expectation of the research is that actions taken to alleviate congestion are appropriate under heavy traffic conditions as described above, and are curative. The objective is to maximize throughput in order to minimize or eliminate residual queues. And if these actions fail, one expects that the objective will shift to minimizing the damage of queue formation to the extent possible, which is palliative. The analysis of the example applications and interview responses will be to determine the accuracy of these expectations.
In this case study, the wide median of the Michigan "boulevard" prevented normal opposing left turns. Left turners first turn right and then U-turn to complete their maneuver. Thus, left turners approach the traffic signal twice.
Before improvements were made, the operation started with westbound green and a green for the signalized U-turn. This provided uninterrupted flow for left turners to first turn right and then make their U-turn. After a period of time, the U-turners were stopped to allow the southbound through movement to proceed up to the intersection. That southbound movement then received a green signal at the intersection.
Finally, the southbound movement was stopped at the U-turn to allow remaining U-turners to make their U-turn and still be able to go through the intersection with a southbound green signal.
In terms of its operational effect, this sequence first allowed left turners to turn right and then U-turn to then store in the southbound intersection approach. The southbound movement at the U-turn then turned green to serve southbound through cars. This condition prevailed for a while until the southbound movement at the intersection received a green. Thus, the operation was intended to provide southbound progression primarily, and depended on relatively small volumes making the right-turn-U-turn sequence of maneuvers. The problem was that the volume of U-turners increased to the point that it filled up the southbound approach at the intersection, which caused the queue to block the progression for southbound through cars. The southbound approach became congested, and the operation favored those making the left-turn maneuver at the expense of the southbound maneuver.
The objective of the before condition was to provide smooth flow for the southbound through movement, but that objective was thwarted by congestion from the oversaturated southbound approach and the queue formation from U-turners.
To minimize the problem, staff at the Road Commission for Oakland County altered the sequence. The altered sequence started by providing an open southbound green at both the U-turn and the intersection. This prevented the queue from the U-turn from blocking southbound through cars. After the southbound movement was served, the sequence served the U-turn.
The objective of the altered operation is to manage queue spillback to prevent the queue from a relatively unimportant movement from blocking a relatively important movement.
The sequence of movements that provided the most effective queue management is shown in Figure 2-3.
Figure 2-3. Sequence of four diagrams showing signal sequence at Square Lake road and Telegraph Road in Oakland County, Michigan. The intersection is an east-west boulevard that requires left-turners to turn right and then make a U-turn on the north-south crossing street. The U-turn is signalized. First interval: Green southbound at U-turn and boulevard. Second: Green westbound boulevard and U-turn. Third: Green for the U-turn and the southound through movement at the boulevard. Fourth sequence is the same as the first, with green for all southbound movements.
This congested corridor extends a little less than three miles through 11 signalized intersections from White Rock Road to Gold Country Boulevard, as shown in Figure 2-4. Traffic volumes on Sunrise increase from 1500 vph at the south end to 5000 vph at the north end in the peak direction. Sunrise Boulevard carries 80,000 vehicles per day on three basic lanes in each direction.
In the PM peak, demand exceeds capacity at the three northernmost intersections. Queue formation extends back from the first of these intersection upstream over the length of the section.
Figure 2-4. Figure showing Sunrise Boulevard in Rancho Cordova, California. The diagram shows volumes steadily increasing along arterial road to a peak of 5000 vehicles per hour at the most downstream of 12 intersections. This intersection is circled, showing the bottlneck where demand exceeds capacity.
The agency has considered widening, but is constrained by limited right of way.
Alternatives considered include:
All but the final approach are curative, seeking to maximize throughput with the hope of minimizing or eliminating congestion.
The flush strategy is currently being used. It uses vehicle preemption to force the signal to pause while serving northbound Sunrise traffic. The flush is implemented manually. Nothing is done to alleviate the resulting congestion on side streets.
The desired strategy currently under consideration is an adaptive flush strategy implemented by time of day. The strategy would be defined by thresholds, possibly queue formation. Minor objectives would be to keep northbound turn bays clear, and to alleviate congestion on some major side streets. The objective is to focus queue formation on the portions of the roadway with the most favorable storage.
Both the manual and contemplated automatic flush strategies are palliative. They seek to move the damaging queues to the least damaging location. There is no expectation of alleviating residual queues, nor is there any expectation of maximizing throughput overall.
The location in question is a network of three signals, two at the ramps of I-40 and NC-54, and one at NC-54 and Farrington Road, just west of the southbound ramp. See Figure 2-5.
Figure 2-5. Figure showing aerial photo of North Caroline State Highway 54, running east-west, and IH-40, which in this location runs north and south, in Durham, North Carolina. The diagram shows a wide diamond interchange with two signalized ramp intersections, and an at-grade signalized intersection immediately to the west of the southbound ramp. The northbound I-40 traffic exiting to go west on NC54 have a loop ramp.
The principle problem is a severe capacity constraint at the NC-54/Farrington intersection, which backs the queue back into the ramp intersections and onto I-40.
The existing operation attempted to provide progression at the start of green using a 153-second cycle length. Excessive queues formed westbound that backed up around the loop ramp to the northbound I-40 mainlanes. Both directions of Farrington Road also suffered residual queuing. Clearly, the objective to provide progression was not appropriate for the oversaturated conditions.
The first attempted corrective measure employed offsets intended to create storage for side-street traffic. This approach involved simultaneous reds on NC-54 so that the side streets would have access to the storage area. A key feature of this alternative was the use of shorter greens and cycle lengths to minimize turn-lane overflow. The green time was set to roughly empty the turn bays, and the red time set to roughly fill the turn bays. The cycle used was 110 seconds. This approach somewhat alleviated queue formation on Farrington, but the queue still backed out onto I-40, and the westbound left turn at Farrington still spilled out of its bay.
The first attempted approach was intended to avoid queue overflow in an attempt to maximize throughput. The hope (unrealized) was a cure for excessive queuing by maximizing throughput. A minor objective was to alleviate excessive delay on Farrington Road.
The second attempted corrective measure sought to assign capacity based on safety needs rather than delay. The alternative lagged the westbound left turn to reduce queue spillback and increased the cycle length to improve capacity. The cycle used was 150 seconds. The simultaneous main-street reds were retained, but the capacity on the Farrington approach was reduced.
The result of the second attempted approach was excessive queuing onto I-40 and excessive queues on Farrington Road.
This approach was still based on maximizing throughput to alleviate or minimize residual queuing. The objective was not met.
The third attempted strategy finally employed a pure queue management objective. The cycle was increased to permit a higher green percentage on westbound NC-54. Splits for movements that had safe storage were reduced. The controller was programmed with holds and recalls to prevent actuation from restoring green time to movements of less importance to the objective of safe queue storage. The cycle used was 200 seconds. The controller was programmed so that plan changes occurred on even multiples of the cycle length, with the same offset used in all plans. This technique eliminated plan transitions at critical intersections.
The result of the third solution was excessive queuing on Farrington Road, but lessening of the queue on westbound NC-54 such that the queue did not spill out into I-40 mainlanes.
None of the solutions provided acceptable operation using any objective other than queue management.
Bandera Road is one of the major arteries serving northwest San Antonio west of IH-10. At the time of this example (early 1990's), the traffic signals along Bandera were coordinated for the first time. The problem was a very heavy outbound left turn movement, as depicted in Figure 2-6. Widening to allow a two-lane left turn was not an option at the time, though doing so would have solved the problem. This case is typical of many, in that improvements to the lane utilization can often double the capacity of congested movements, while finding more green time is usually a marginal improvement.
Figure 2-6. Figure showing divided highway (Bandera Road) intersecting a minor side street (Guilbeau Road) in San Antonio, Texas. The figure highlights the heavy left turn of 950 vehicles per hour from westbound Bandera Road to southbound Guilbeau Road.
The existing traffic signal phasing was conventional split phasing, given that most traffic on the side streets turned left to enter the main highway. The time required to serve the side-street movements exclusively consumed too large a share of the cycle, and the result was severe congestion on the outbound left turn and a growing residual queue that exceed a mile in length. When the queue spilled out of the left turn lane, it blocked a through lane, which caused congestion on the through movement.
The left turn volumes on the side street were light, and the first operational improvement was to eliminate the split phasing and require side-street left turners to yield. This resulted in a high level of public complaint, which compelled a different solution.
The next solution used phase reservice, which is a technique that appears over and over in discussions with expert practitioners, as described in the next section. Serving phases twice in the cycle may be used to reduce the number of times a minor movement is served, or increase the number of times a major movement is served, relative to the coordination. In this case, the intersection was operated at twice the normal coordination cycle of the system, and the main-street movements were served twice in the cycle. Thus, the effective cycle of the main street movements was consistent with system timing, but the percentage of that effective cycle consumed by the side street was significantly reduced.
Figure 2-7 shows the final phasing scheme.
Figure 2-7. Phase diagram for Bandera and Guilbeau Roads. The controller 8 sequential phases: 1, 2, 5, 6, 3, 4, 7 and 8. Overlaps are: West thru: 2+4+5+7. West left:1+2+3+4. East thru: 5+7. Eastbound left: 1+3. South is Phase 6, and north is 8. Phases served: 1. east and west lefts, 2. west thru and left, 5. east and west thru, 6. southbound, 3. same as 1, 4. same as 2, 7. same as 5, 8. northbound. Phases 3, 4, and 7 can be permitted to serve main street twice in cycle, or omitted for conventional split phase operation.
It should be noted that since this operation was implemented, the intersection was widened to provide a two-lane left turn. This operation, however, improved throughput until geometric improvements could be made.
Interviews were conducted with a variety of recognized experts. These experts are listed alphabetically below:
The results reported below also include insights gained by the author while serving with the City of San Antonio in Texas.
A rough outline of questions was used as a means of promoting discussion. These are shown below:
In the actual interviews, the discussion ranged away from the questions significantly. This was allowed, because the objective was not to prejudge the knowledge of the respondents with leading questions, but to discover that knowledge in its native organization. Therefore, the questions provided a rough outline of points that needed to be included in the discussion. It was not productive, however, to attempt to categorize the interview responses according to the above questions. Therefore the responses summarized below are organized by themes that emerged during the interviews.
The definitions offered by the interviewees varied in terms of specificity. These are some samples:
All of these are related. In all cases, the observed phenomenon is growing residual queues. In these cases, delay cannot be evaluated. All define the saturated condition after making as many improvements as possible to green splits at the intersection. Most noted that green splits may be dictated by constraints unrelated to congestion on the approach.
With this definition, the first strategy was first to make as many conventional green-split improvements as possible before declaring the intersection as exceeding saturation. This corresponds to initial curative treatments, where the hope is to relieve the congestion altogether. The objective based on these definitions is either maximizing throughput or minimizing damaging queues.
Practitioners are usually stretched thin enough such that they require conditions to deteriorate to an action point before they are compelled to do something. The respondents varied in the conditions that would compel action. Several respondents took action when observing operation during routine reviews of new signal timings. All the respondents agreed that observing an unexpected residual queue would compel action, especially if it grew to the point of causing gridlock or a safety problem, and one mentioned accident experience as a motivator. Most reported that citizen complaints were often the trigger that compelled action. One responded to requests for assistance from state DOT district offices, where citizen complaints were received.
The first step for all respondents, after confirming that there was a problem, was to first adjust the green splits to minimize residual queuing. When this didn't work, the respondent would determine that the intersection was oversaturated and start to consider more specialized strategies. There was no particular sequence to the strategies. Some respondents did not distinguish between the objectives of maximizing throughput and managing queues, while others specifically changed strategies when it was no longer possible to provide sufficient throughput to avoid damaging queues. Thus, some strategies are oriented towards maximizing throughput in the hopes of minimizing damaging queues, while other strategies were reserved for situations where that objective could not be reached. We should note, however, that even though these strategies are divided into the two groups, there is some crossover between them. Most respondents did not make the distinction between throughput and queue management.
We should also note that none of the respondents used optimization tools currently available as part of designing or implementing these strategies, and only two used any form of simulation as a means of evaluating the strategies. All respondents felt that each situation was unique enough to defy the use of a specific sequence of analysis steps. All respondents also felt that improving the problems required the sum of multiple small improvements. Finally, all respondents agreed that the problems and solutions were best identified during direct field observation.
Throughput maximization strategies included:
Queue management strategies included:
One respondent presented a strategic concept that bears special mention. He made the distinction between two general strategies:
One respondent also presented a series of "demons". The demons were those influences that forced signal timing into states less able to address congestion. These included:
The respondents used a range of system tools and controller features to support their objectives at saturated intersections. One respondent specifically avoided the use of esoteric features as a matter of policy, because of the concern for maintaining the implementation of those features with maintenance forces who don’t understand them. That leads to a general recommendation from all the respondents: All specialized controller features used in congested conditions should be defined as straightforward features, rather than a combination of otherwise unrelated special features that is hard to comprehend and maintain. One example is a straightforward feature for serving a phase more than once in a cycle. There are many approaches to doing this (including the method shown in Example 4), but if some agencies are reluctant to overwhelm their technicians with the use of such features, then the solution is to make the features available without overwhelming the technicians.
Overriding SCATS Control. One respondent had a SCATS system running the signals in his jurisdiction, and that provided a means of metering traffic into congested intersections. When the queue from a congested intersection backed through an upstream intersection, the normal SCATS response was to increase green time on the congested approach. Given that the congestion was caused by the downstream intersection, the result was wasted green time, with crossing movements left unserved. He used a special detector to sense the queue approaching the upstream intersection, which triggered SCATS to reduce rather than increase traffic on the through movement feeding the congestion. This would allow minor movements to be served.
The respondent's experience with SCATS brings up a more general issue of adaptive control in saturated conditions. The objective of current adaptive control systems is to adjust timing so that the critical intersection hovers at a high level of saturation, typically about 90%. Adaptive systems respond to the intense detection response associated with a standing queue by adding more time to that movement. They do not balance queue lengths, nor do they have any mechanism for addressing damaging queues. When demand increases, they all seek longer cycles, even though this may not be the preferred strategy to maximize throughput. As with off-line optimization software, none of the adaptive optimization algorithms in common use seek to maximize throughput or minimize damaging queues as an objective function. For this reason, the respondents generally did not consider adaptive control as likely to be effective in congested conditions, particularly when considering single or small groupings of intersections. Only one respondent manages an adaptive system routinely, and as the above paragraph explains, he sometimes has to work around that system's algorithm.
Proponents of adaptive control suggest that these systems may delay the onset of congestion even if they are less effective once the congestion arrives. As a matter of opinion on the part of the researchers, this varies depending on the adaptive system being considered. Systems that minimize queue formation (as a byproduct of maximizing throughput) will prevent congestion most effectively, and those considering adaptive control would be encouraged to understand the objectives of the optimization process used by the adaptive system.
Phase Reservice. Several respondents manipulated their controllers to provide green signals more than once in the cycle, or to alternate movements to every other cycle. For example, the author once used a cycle twice the coordination cycle to serve split-phase side-street movements on alternating cycles. These movements were light, and serving them on alternate cycles made better use of the minimum green time, which was controlled by pedestrian clearance time.
One respondent suggested that the more heavily movements are imbalanced, the more likely the intersection will benefit from serving the major movement more often in the cycle, for shorter periods.
Minimize Pedestrian Effects. When pedestrian demand is heavy, then all respondents agreed that it had to be included in the signal timing explicitly. But when pedestrian demand is light, the temptation by many agencies is to avoid installing pedestrian signals and pushbuttons. As mentioned above, the presence of signals and pushbuttons allows pedestrian movements to be routinely ignored unless there is a pedestrian call. The traditional approach to this is to provide coordination timing that violates the pedestrian intervals. When the pedestrian intervals are called by a pushbutton actuation, they override the coordination timing to serve the pedestrians. The intersection then allows the signal to transition back to coordinated operation. If this happens only rarely, the loss of coordination has negligible overall effect. In this case, "rare actuation" is considered to mean an actuation no more often than once in several hours.
Not all controllers handle pedestrian override as defined above. For example, some controllers transition slowly, while others won't allow coordination timings that violate pedestrian intervals to be installed in the first place. One respondent suggested a workaround. He sets the split to allow for the pedestrian intervals, but then sets the maximum green time (MAX) in use during that pattern at the normal green time needed by vehicles. The phase will normally max when the pedestrian intervals are not called. If all the MAX times are set to their desired vehicle splits, then the MAX times will control the splits of the non-coordinated phases. Some controllers inhibit MAX during coordination, and this must be disabled. When the pedestrian intervals are called, the intersection runs according to the coordination timings, with no loss of coordination and no requirement for transition.
Some controllers allow MAX settings by pattern, and some by time of day.
Detector Switching. Several respondents used detector switching to turn off detectors that cause an approach to extend for density values that are too low, and for turning on detectors that serve greens that overlap with other movements to increase efficiency. An example is to switch off a detector in a right-turn bay if the complementary left turn will be served next (i.e., if it has a call). This requires using special controller logic.
Special Controller Logic. One respondent aggressively used auxiliary logic functions within controllers to achieve specific objectives. For example, when a left turn lags, he uses an auxiliary function to extend the complementary through movement even if it has no call.
Another respondent used controller logic to circumvent normal barrier cells. The logic controls in their software allow them to assign a logic output to a load switch, and then control that logic output by any combination of normal controller conditions.
Simultaneous Gaps. Once one of a lagging phase pair gaps, the controller will not extend that phase until the next cycle, even if it is holding in green waiting for the other lagging phase pair to terminate. This prevents alternating widely spaced cars to hold opposing through movements in green despite low density being served.
Dual Entry. This feature causes compatible minor movements to both turn green even if only one of the phases was called. For example, if the side street has a call on Phase 8, the controller will also serve Phase 4 along with Phase 8. Without invoking this feature, Phase 8 would be served while Phase 4 would remain red, unless a car placed a call some time later in the Phase 8 green period, at which time Phase 4 would go green and time all its minimum intervals. Dual entry forces side-street compatible pairs to be served at the same time.
Dynamic MAX. One respondent mentioned this feature, though it was not favored by all respondents. If a phase maxes out in successive cycles, the MAX time will be increased by an increment. This will continue for each successive maxing out until a maximum MAX is reached. Some controllers revert the max time back to the normal setting on the first phase that gaps out instead of maxing out, while others will subtract increments on successive gap-outs until it backs down to the normal setting.
A variation on Dynamic MAX is adaptive split operation. One respondent reported the desire to have coordination timings adapt to the presence of detection throughout the green period. In other words, when cars are detected right through to the end of green repeatedly, the controller would adjust the coordination timings accordingly.
Volume-Density. Use variable gap to shorten gap settings quickly to more aggressively find a gap in the approach traffic. Also, use variable gap on stop-line detection zones to prevent a truck-induced (and therefore larger) gap from causing a short green.
Most of the desired controller features are available from one or two manufacturers, but they are not part of the NEMA standard and therefore not universally provided. Some require logical functions external to the basic NEMA ring rules. Desired controller features include?
United States Department of Transportation - Federal Highway Administration