Bus-Friendly Traffic Signals Can Reduce Bus Delay By 90% on South Huntington Avenue


At two intersections on Boston’s South Huntington Avenue – intersections with Perkins St. and Bynner St. (Figure 1) – buses on Route 39, the MBTA’s fourth-busiest bus route, experience an average combined delay of 43 s. This post will explain how, with standard technology and bus-friendly signal timing, bus delay through this pair of intersections can be reduced to only 5 s. These changes will also reduce pedestrian delay by more than 60% (and even auto delay declines a bit). All this without removing parking lanes or bike lanes.

Figure 1. South Huntington Ave, with intersections of interest and bus stops indicated. Base map: Google Maps, 2024.

Transit Signal Priority

One key to reducing bus delay is to apply transit signal priority (TSP), which means detecting buses as they approach a signalized intersection and adjusting the traffic signals so that buses can pass through with little or no delay. Technology to apply TSP has been around since the late 1970’s; in the last 15 years, TSP has become even easier to apply, using GPS to detect buses and and short-range radio frequency for communication. Widely used in Europe for 40 years, it is still relatively uncommon in the US.

At the two South Huntington intersections, the only TSP tactic of practical interest is Green Extension, which means that if a bus is detected while the signal is green, the signal will be held in green until the bus passes (Figure 2). How effective Green Extension is depends mainly on:

  • How long an extension can be granted? This is limited by the prediction horizon – how far before the intersection can bus arrival time be predicted, which is the travel time from the previous bus stop; and that, in turn, depends mostly on whether the bus stop is on the near side or far side of the intersection. As Figure 1 shows, at the two intersections being studied, there are two far-side stops and two near-side stops. For the two far-side stops, the prediction horizon is 20 s, while for the near-side stops, it’s only 3 to 5 seconds, so TSP can do very little there.
  • How long is the bus street red? Depending on the intersection and the period (a.m. or p.m.), the red for South Huntington Avenue currently lasts between 60 and 90 seconds.

Figure 2. Illustration of Green Extension

The difference between the bus street red and the prediction horizon is the bus effective red, which is the main factor that explains bus delay. For example, at Perkins Street in the a.m. peak, the red time for S Huntington Ave is 80 s, and so northbound, with a far-side stop and therefore a 20-s Green Extension, buses could still experience a 60 s delay. Southbound, because the bus stop is near-side stop and so the green extension can’t be longer than 5 s, buses could still be delayed by 75 s.

So – TSP can reduce bus delay a little, but bus delay could be reduced a lot more if the bus street red were shorter, and if all the stops were far side.

TSP-Friendly Signal Control

Because of the long bus-street red and the limited detection horizon, layering TSP on top of the existing signal control will have only limited effectiveness. To make it more effective, can the underlying signal timing be changed so that Huntington Ave has a shorter red? And because the length of the red interval is closely tied to the length of the signal cycle, this question becomes: can the underlying signal timing be changed to have a shorter cycle?

Yes. Using standard intersection capacity analysis and some clever signal cycle design, we found that the traffic and pedestrian needs of these intersections can be met with cycles ranging from 40 to 70 s, a lot shorter than the cycles that operate today in the range of 100 to 134 s (see Figure 3). The most dramatic reduction is at Perkins Street in the p.m. peak – instead of 134 s, the cycle length can be just 40 s, with bus red falling from 90 to 20 s.

Figure 3. Existing and Proposed Red Times, Green Times, and Cycle Lengths

Peak hour turning movement counts for the intersections, taken from the City’s database, are shown in Figure 4 to Figure 7. The counts are from June, 2015, and based on our field observations in 2023-24, closely match traffic patterns today. It is noteworthy that while most turning flows are small, there are a few heavy turning flows:

  • in the a.m. peak, from Perkins and Bynner eastbound, turning left onto South Huntington
  • in the p.m. peak, from South Huntington southbound, turning right onto Bynner and, to a lesser degree, turning right onto Perkins (the opposite of the a.m. peak heavy turns)

Figure 4. Perkins St AM Peak Turning Movement Count

Figure 5. Perkins St PM Peak Turning Movement Count

Figure 6. Bynner St AM Peak Turning Movement Count

Figure 7. Bynner St PM Peak Turning Movement Count

The key to enabling short cycles is making the pedestrian crossings concurrent with the parallel vehicle phase instead of making pedestrians wait for an exclusive pedestrian phase (EPP). Because concurrent crossings usually involve “permitted conflicts” with turning traffic (advancing on a green ball, with an obligation to yield to crossing pedestrians), they should only be used where those conflicts can be managed in way that makes the crossings safe. To that end, the City of Boston’s Signal Operations Design Policy has the following limits on the number of permitted turning conflicts crossing pedestrians should have to face:

  • Left turns across a single lane: No more than 2.5 permitted conflicts per cycle.
  • Right turns:
    • With an LPI: No more than 5.5 right turns per cycle
    • Without an LPI: No more than 3.5 right turns per cycle

A Leading Pedestrian Interval (LPI) is a pedestrian head start, usually lasting 4 seconds, that enables pedestrians to establish their priority in the crosswalk before parallel traffic is released.

While currently there are more vehicles turning per cycle than those limits would allow, with a shorter cycle, there will be fewer turning conflicts per cycle – for example, at Bynner Street in the p.m. peak, with a cycle length of 110 s, there are currently 7.2 right turns per cycle; but with the 55 s cycle we propose, there will be only 3.6 right turns per cycle.

Existing and proposed phasing plans (they apply at both intersections) are shown in Figure 8. One can see that the existing plan has four stages, including an exclusive pedestrian phase (Stage 4), while the proposed plans have fewer stages and concurrent crossings. In the phasing diagrams, high volume turns are indicated while low-volume turns, which will run with the thru movement, are suppressed; and a solid turn arrow means the movement is “protected” (operates conflict-free under a green arrow), while a dashed arrow indicates a “permitted conflict,” meaning a turning movement running under a green ball, with the obligation to yield to parallel traffic and/or pedestrians.

The proposed plans were designed to more than satisfy Boston’s limits on permitted conflicts. For the a.m., the challenge of high eastbound left volumes was met by making Stage 1 (a leading thru-left phase) long enough that most of the left-turning traffic from Perkins Street eastbound will pass during Stage 1, so that no more than 2.0 left turns per cycle occur during Stage 2 (the stage with the permitted conflict). Further adding to pedestrian safety is that during the early part of Stage 2, when pedestrians begin crossing the intersection’s northern leg, conflicting left turns will be blocked by the just-released westbound thru traffic.

Figure 8. Stage Diagram for Both Intersections

For the p.m. peak, there is no need for a leading left phase because of low left turn volumes, and so the proposed phasing plan has only two stages. However, because of high right turn volumes from South Huntington, Stage 2 begins with a Leading Pedestrian Interval (LPI) as Stage 2A, lasting 4 seconds. Thanks to the short cycle lengths (40 s at Perkins, 55 s at Bynner), the number of right-turning vehicles per cycle during Stage 2B will be only 1.0 at Perkins and 3.6 at Bynner. These volumes are well below the City’s limits; furthermore, due to the tight intersection geometry, most right turns there are made at a speed of around 10 mph.

Detailed timing plans are shown in Figure 9. The phase times (“splits”) are long enough to serve the concurrent pedestrian crossing with a 7 s Walk interal and a clearance time designed for a pedestrian walking speed of 3.5 ft/s (those are standard values). The pedestrian crossings will be on recall, i.e., automatic.

Figure 9. Proposed Timing Plans

TSP-Friendly Intersection Layout

For TSP to be effective, the intersection layout needs minor changes to support bus priority. First, as suggested earlier, move the two near-side stops to the far side (see Figure 1). TSP can only be effective with far-side stops because they allow the advance notice necessary for Green Extension. A field visit confirms that the two needed relocations are feasible.

A second needed layout change is to add short left-turn pockets to the South Huntington Avenue approaches. While few vehicles turn left from South Huntington – none of those left turn movements has more than 1.2 vehicles per minute – when a vehicle turns left, they block through vehicles, including buses. We observed this to be one of the causes of long bus delays.

The left turn pockets can be very short – long enough to hold 1 car so that buses and other thru traffic can bypass it, as shown in  Figure 10. Short turn pockets like this can be added without affecting the protected bike lanes, and only one or two parking spaces will be lost.

Figure 10. Proposed layout changes: far-side bus stops and a very short left turn lane


To test how effective these changes would be, traffic in the corridor was simulated using PTV Vissim. Three alternatives were tested: existing, existing with TSP overlayed, and the proposed timing plan with TSP. The final alternative also includes the proposed TSP-friendly intersection layout changes. Here are the results:

Bus Delay: Figure 11 shows that while adding TSP alone reduces bus delay to some extent, TSP coupled with TSP-friendly traffic signal timing and a TSP-friendly intersection layout can make bus operations here almost free from traffic delay, with average total delay through the two intersections falling from 43 s to less than 5 s.

Figure 11. Bus Corridor Delay

Pedestrian delay: Figure 12 shows that pedestrian delay, averaged over all crosswalks, falls dramatically, from 47 s to 18 s, with the proposed timing and TSP. With concurrent crossings and such low average delay, pedestrian compliance is bound to improve substantially. Notice that if TSP is merely overlaid on the current timing plan, pedestrian delay actually worsens.

Figure 12. Pedestrian Delay

Auto delay: Figure 13 shows the impact on vehicle delay, averaged over all vehicles that enter the simulation zone (some pass through one intersection, some through two). Simply adding TSP increases auto delay a little. However, because the proposed timing plan is so much more efficient than the current plan, auto delay falls a lot, even though buses have the power to interrupt the signal cycle for up to 20 s.

Figure 13. Auto Network Delay


NatDave Obeng-Amoako, a PhD student in Northeastern University’s Transportation Engineering program, did this study together with his adviser, Peter Furth.

APPENDIX I:  A New Traffic Signal Cycle Design Method for a Protected Plus Permitted Left Turn Phasing with a Limit on Permitted Conflicts with Pedestrians

As part of this study, we developed a new method for designing signal cycles that involve protected plus permitted left turn phasing for one approach. The innovation is that this method limits the number of vehicles expected to turn left during the permitted interval, when pedestrians will be crossing concurrently. On streets that have left turn lanes, it is safer – for both pedestrians and vehicle occupants – to have protected-only left turns. However, on roads like Perkins and Bynner Street that are too narrow to have left turn lanes, protected-permitted phasing with limited permitted conflicts offers an efficient yet safe alternative to exclusive pedestrian phases and split phasing, two ways of avoiding permitted turn conflicts with left turns in the absence of left turn lanes.

For this exposition, the phasing plan shown in Figure 8 is assumed, reproduced here as Figure 14, with left turn conflicts being a concern for only one approach (in Figure 14, the eastbound direction), and assuming that the duration of Stage 2 (the permitted left turn phase) is governed by pedestrian timing needs rather than vehicle capacity needs. The Stage 2 pedestrian phase is assumed to be on recall (i.e., automatic).

Figure 14. Protected-Permitted Left Turn Phasing

Needed Length of the Protected Left Turn Phase

Appendix II: Traffic Signal Coordination Does Not Help Buses

On corridors with closely-spaced intersections (quarter mile or less), signals are sometimes coordinated to enable vehicles to pass through a series of traffic signals without stopping. We tested coordinating the two signals for buses, but it turned out that because of how frequent the Route 39 service is, coordination would actually worsen delay for buses (and cars) because once a bus interrupts the cycle, the signals often don’t have time to get back in sync before another bus arrives.