2-Second Green Shifts: Adaptive Signals Beat Lane Additions

A travel-time saving can be locked inside the signal controller cabinet, not in the next lane-expansion project. Adaptive signal control reads incoming traffic and reassigns green time in real time, so a corridor can make use of unused cycles that fixed-time plans throw away every peak period. That is the central trade-off: adding lane miles costs money, while adding adaptive logic costs a fraction and produces the same directional benefit.

Fixed-time signal plans run on historical timing schedules, so they cannot respond when a side street empties early or an arterial's through-movement backs up. Adaptive control closes that gap by extending or shortening phases based on live detector data. The result is a better return on the existing street network: the same lanes, the same intersections, but with green time allocated where it is actually needed.

Before committing bond funds to pavement, agencies should quantify what a recovery in travel time means for their own corridors. The signal cabinet already contains the cheapest lane miles a city will ever build — they just happen to be measured in seconds rather than in asphalt.

The Green Shift

SCATS moves green time in small increments. That granularity, not the full signal cycle or the corridor plan, is the real unit of adaptive signal control (ASC). ASC is an optimization layer that sits on top of the local controller: every few seconds it reallocates green split, cycle length, and offset using current per-approach demand rather than a time-of-day schedule. The local controller still handles lamp switching and detection; ASC just changes its instructions far faster than an engineer ever could.

The core tuning unit is the sub-phase green shift. SCATS adjusts green split in small increments and cycle length in slightly larger increments, so a brief gap in one approach — a slow truck, a pedestrian who cleared early — is reassigned to another approach before the arriving wave gets there. Time-of-day plans assume demand at one time of day looks like demand at another; ASC assumes nothing about the immediate future and re-checks constantly.

Surtrac, developed at Carnegie Mellon University, pushes the interval further. It re-optimizes frequently using predicted arrival profiles, not just detector presence at the stop bar. In a Pittsburgh deployment, Surtrac cut travel times substantially, a result published by the CMU Robotics Institute. "Predicted" is the load-bearing word: the optimizer anticipates the platoon that will arrive shortly rather than reacting to cars already sitting at the light.

Downstream coordination runs on those predictions. Each controller sends the next signal an expected arrival time for the platoon it released, and the downstream optimizer treats that time as a hard constraint. Offsets become reactive rather than preset. In conventional coordinated operation, offsets are fixed in the field; under ASC, an offset is a continuously updated estimate of when a specific group of vehicles will reach the next light.

The safety valve is a degree-of-saturation cap. ASC algorithms limit each approach to high saturation, so green extensions stop before the downstream link can overflow. That rule is what prevents an adaptive system from simply shoving the queue one intersection down the road, the standard failure mode of naive retiming.

None of this requires connected vehicles, AI forecasting, or a new traffic-management center. The Provo, Utah deployment demonstrated that with modern controllers, existing detection, and software that reallocates green seconds in real time. The software layer is the investment, not the hardware.

That is why the green shift beats the lane adder. Lane widening adds link capacity but leaves phase timing untouched; ASC changes timing at the same granularity at which congestion forms. For a corridor whose top signals run hot at peak, the situation quantified elsewhere in this guide, the first move is the optimization layer, not the asphalt.

The Evidence File

Before the Provo deployment gets dismissed as a single-city outlier, several independent before/after evaluations document the same mechanism with conventional detection only. The strongest evidence for adaptive signal control is not any one headline number; it is that travel-time reductions appear consistently without connected-vehicle infrastructure, AI forecasting, or a new traffic-management center.

Bellevue, Washington, supplies the cleanest mid-sized-city case. According to a city performance memo, Bellevue placed many signals under adaptive control and measured PM-peak corridor travel times with Bluetooth probe runs before and after. The corridor run fell — a reduction. The deployment used existing detection and modern controllers, not a greenfield technology build.

Colorado's US-36 corridor provides the off-peak control condition most before/after evaluations lack. According to a Federal Highway Administration before/after report, CDOT placed several intersections under adaptive control and measured a peak-period travel-time reduction; off-peak travel time did not change significantly. That asymmetry is mechanistically important: adaptive control reallocates green seconds where queues form at peak, and it correctly does little when the queue is absent.

Those two results sit inside a broader distribution. According to FHWA's ITS-JPO synthesis of adaptive-control sites published previously, the median travel-time reduction fell in the middle of a documented distribution. The corridor-level headline used in this guide is not the ceiling; it is the upper-middle of that distribution, which makes it a defensible planning assumption rather than a best-case sales figure.

According to a Texas A&M Transportation Institute evaluation of Texas arterials running InSync video-based adaptive control, stops fell and average travel time fell as well. Stops matter more than travel time for the payback math: fewer stops means less brake wear, less idle fuel, and shorter residual queues at the downstream intersection — which is why corridor-level benefits compound beyond raw minute savings.

Consistency check: every evaluation above used conventional signal detection — inductive loops, video, and Bluetooth probes — and none required connected-vehicle infrastructure. That is exactly why the Provo result is not a one-off. The detection layer is already in the ground in most mid-sized cities; the marginal investment is software and controller upgrades, not a new sensor network.

One edge case before you scale the planning value: the US-36 off-peak finding implies that corridors with genuine around-the-clock congestion — not intersection-bound peak delay — will see smaller relative gains. If your top signals clear the peak-delay threshold but off-peak delay is also degrading, adaptive control still helps, but shade the corridor-level assumption down.

For a city team deciding what to cite: Bellevue is the peer analog, US-36 is the mechanism confirmation, and the ITS-JPO synthesis is the planning-range source. When most of a corridor's top signals show peak-hour intersection delay above the threshold, this evidence supports adaptive control before lane-widening, with the reduction as a realistic planning value and sensor maintenance funded for the long term.

Cost Decision

The schedule gap changes the funding question, not just the calendar. An adaptive retrofit can be delivered inside one budget cycle — meaning the headline corridor-level reduction covered earlier in this guide starts paying before a lane-widening project clears right-of-way. Lane widening needs several years for right-of-way, environmental review, and utility relocation, forcing a bond-scale capital decision and multi-term political risk. Fixed-time retiming is quicker still, but its return decays as traffic patterns shift, so the city re-pays whenever the peak moves.

The risk row inverts the usual intuition. Adaptive's failure mode is a graceful fallback to fixed timing — safe, unimpressive, and no worse than the pre-project state. Lane widening's failure mode is induced demand: VTPI attributes a vehicle-miles-traveled increase per added lane-mile in growing metros, meaning the new capacity is partially consumed by the traffic it attracts. The cheap option's downside is the status quo; the expensive option's downside is a more congested road with more vehicles on it.

The decision rule that follows: when most of a corridor's top signals show peak-hour intersection delay above the threshold, choose adaptive signal control before any lane-widening project, and fund the sensor maintenance line for a long-term horizon. The comparison table makes the cost case.

Currently, the corridor-level before/after numbers in the ASC literature are real, but they are not the number your city will experience. The early ASC evaluation pool was selected from corridors with obviously poor fixed timing; an agency that hasn’t touched its signal plans in years is the ideal experiment. Cities that already retime regularly typically see only a modest additional gain from going adaptive, because the low-hanging green-time misallocations are already gone. The headline corridor gain is therefore not a guaranteed baseline — it is an upper bound for a badly tuned incumbent system.

The mechanism can also move delay rather than reduce it. A California PATH evaluation of SCATS on Oakland’s Broadway corridor found peak delay increased, because Bay Bridge spillback left the downstream signal saturated. Adaptive control can reallocate green time and shift queues, but it cannot create downstream storage. If the bottleneck is outside the corridor’s boundary — a bridge approach, a rail crossing, a major merge — the controller is managing a queue it cannot discharge.

What the Data Doesn't Tell You

The second hidden assumption is that the detection layer works. An ITS maintenance audit in Tennessee found many signal loop detectors were nonfunctional. When a loop fails, ASC silently falls back to fixed timing; no alarm sounds, no email is sent, it just runs the stored plan. The reported savings therefore assume an operations-and-maintenance budget many cities do not actually fund. A long-term sensor-maintenance line item is not a bureaucratic add-on; it is the control system itself.

The measurement denominator does a lot of work, too. Most before/after evaluations report corridor-level, peak-direction travel time. Include off-peak, weekends, and cross-trips, and the citywide person-minute savings drop to a small share of total commute delay, even when the corridor number looks large. For a mid-sized city, this is still a good return on a cheap retrofit — but it changes how you explain the project to a council expecting citywide congestion to disappear.

The mean also masks variance. The FHWA dataset that includes a travel-time increase site is rarely cited in the marketing slide; installers’ before/after windows are often short, so seasonal and incident effects can swing results. A corridor evaluated after a winter storm will look different from the same corridor evaluated in a dry month.

None of this inverts the decision rule. It sharpens it: adaptive signal control is the higher-return, lower-cost first investment when a corridor’s delay is intersection-based, when the downstream network has room to absorb shifted queues, and when the city is willing to fund detector maintenance for the life of the system. Under those conditions, don’t widen the lane first. But treat every headline number as a conditional measurement, not a physical constant — and check the parameters that made the headline possible before you cut the check.

The takeaway for a mid-sized city: when your corridor meets the decision rule, Provo's sequence is the one to copy — adaptive retrofit first, sensor maintenance funded for the long term, lane expansion reconsidered only if the ASC ceiling is hit.

First, pull the signal timing report, not the travel demand model. The first cut for choosing adaptive signal control over lane expansion is a delay screen on the corridor’s worst signals, measured in seconds of peak-hour intersection delay. If most average above the delay threshold in the target peak, approve an adaptive feasibility study before any widening request is scoped. The threshold is where the signal’s own timing plan, not approach geometry, starts manufacturing delay: the controller holds green seconds on movements that do not need them, and software can reallocate those seconds in real time.

Worked Case

The screen in Rule 1 requires nothing exotic. Currently, none of the rules below assume connected vehicles, AI forecasting, or a new traffic-management center; the Provo gains, detailed in the Worked Case, came from modern controllers, existing detection, and software that reallocates green seconds. If a feasibility-study scope demands any of the first three, the scope is wrong before data collection begins.

Rule 3 sequences the work. Before including a signal in the adaptive corridor, check the downstream volume/capacity ratio for the target peak. If any downstream signal sits at a high volume/capacity ratio, it is oversaturated and its spillback, not signal timing, caps corridor throughput. Fund queue storage or reconfiguration there first, and keep that segment outside the adaptive boundary until the spillback cap is removed. Adaptive green time upstream of a saturated signal does not create throughput; it changes where the queue parks.

Rule 4 is a tripwire written into the contract. Require a long before/after travel-time study with enough probe runs per direction, and specify the decision rule before the first run: if measured travel-time benefit is below a small threshold, revert to updated fixed-time timing and close the adaptive experiment. The long window catches school, holiday, and construction seasonality that a short pilot misses; enough runs per direction gives a stable mean without a massive probe fleet. A below-threshold result means the delay was not primarily intersection-control delay, and Rule 1 should have caught it earlier.

Rule 5 is geographic concentration: choose a set of adjacent signals on a single corridor, not scattered intersections. Offsets require adjacency; a progression needs a common cycle length and a defined spacing between consecutive signals. If the worst signals are spread across several arterials, there is no corridor, only a list. The headline corridor-level results came from networks where every signal in the sequence was included — including the mediocre ones. Skipping one signal to save money severs the offset chain and costs more in lost benefit than the signal’s retrofit would have cost.

Apply the rules in order — screen, gate, sequence, tripwire, concentrate — and the adaptive-vs-widening decision reduces to a memo, not a multi-year modeling exercise. Rule 1 decides whether the congestion is intersection-based; Rule 2 decides whether the city can afford to fix it; Rule 3 sets the corridor boundary; Rule 4 sets the exit; Rule 5 sets the shape that produces the corridor-level benefit. That is the whole first cut.

The takeaway for a mid-sized city: when your corridor meets the decision rule, Provo's sequence is the one to copy — adaptive retrofit first, sensor maintenance funded for the long term, lane expansion reconsidered only if the ASC ceiling is hit.

How to Choose Well: Five Rules for the First Cut

First, pull the signal timing report, not the travel demand model. The first cut for choosing adaptive signal control over lane expansion is a delay screen on the corridor’s worst signals, measured in seconds of peak-hour intersection delay. If most average above the delay threshold in the target peak, approve an adaptive feasibility study before any widening request is scoped. The threshold is where the signal’s own timing plan, not approach geometry, starts manufacturing delay: the controller holds green seconds on movements that do not need them, and software can reallocate those seconds in real time.

The screen in Rule 1 requires nothing exotic. Currently, none of the rules below assume connected vehicles, AI forecasting, or a new traffic-management center; the Provo gains, detailed in the Worked Case, came from modern controllers, existing detection, and software that reallocates green seconds. If a feasibility-study scope demands any of the first three, the scope is wrong before data collection begins.

Rule 2 is the maintenance gate, and it eliminates most adaptive proposals before procurement. Reject adaptive unless the city commits an adequate per-signal annual operating budget for a long-term horizon — the operating cost of keeping detection, controllers, and communication links accurate enough for the optimizer to trust its inputs. If that line item is not in the capital plan, choose fixed-time retiming instead. Fixed-time is cheaper to fail: the update cost is a fraction of an adaptive retrofit, and a poor fixed-time plan still produces repeatable offsets that can be revised on a multi-year cycle. Adaptive without funded long-term maintenance drifts back to fixed-time behavior — after spending the adaptive budget.

Rule 3 sequences the work. Before including a signal in the adaptive corridor, check the downstream volume/capacity ratio for the target peak. If any downstream signal sits at a high volume/capacity ratio, it is oversaturated and its spillback, not signal timing, caps corridor throughput. Fund queue storage or reconfiguration there first, and keep that segment outside the adaptive boundary until the spillback cap is removed. Adaptive green time upstream of a saturated signal does not create throughput; it changes where the queue parks.

Rule 4 is a tripwire written into the contract. Require a long before/after travel-time study with enough probe runs per direction, and specify the decision rule before the first run: if measured travel-time benefit is below a small threshold, revert to updated fixed-time timing and close the adaptive experiment. The long window catches school, holiday, and construction seasonality that a short pilot misses; enough runs per direction gives a stable mean without a massive probe fleet. A below-threshold result means the delay was not primarily intersection-control delay, and Rule 1 should have caught it earlier.

Rule 5 is geographic concentration: choose a set of adjacent signals on a single corridor, not scattered intersections. Offsets require adjacency; a progression needs a common cycle length and a defined spacing between consecutive signals. If the worst signals are spread across several arterials, there is no corridor, only a list. The headline corridor-level results came from networks where every signal in the sequence was included — including the mediocre ones. Skipping one signal to save money severs the offset chain and costs more in lost benefit than the signal’s retrofit would have cost.

Frequently Asked Questions

What does the degree-of-saturation cap do in adaptive signal control?

ASC algorithms limit each approach to high saturation, so green extensions stop before the downstream link can overflow.

What did the US-36 before/after study show about off-peak travel times?

CDOT placed several intersections under adaptive control and measured a peak-period travel-time reduction; off-peak travel time did not change significantly.

How should the corridor-level travel-time headline from the FHWA synthesis be interpreted for planning?

The corridor-level headline used in this guide is not the ceiling; it is the upper-middle of that distribution, which makes it a defensible planning assumption rather than a best-case sales figure.

How does Surtrac differ from SCATS in what it uses to re-optimize?

Surtrac re-optimizes frequently using predicted arrival profiles, not just detector presence at the stop bar.

What should an agency do if its corridor has around-the-clock congestion rather than only peak intersection delay?

If your top signals clear the peak-delay threshold but off-peak delay is also degrading, adaptive control still helps, but shade the corridor-level assumption down.

Why did the Texas A&M evaluation say stops matter more than travel time for payback math?

Fewer stops means less brake wear, less idle fuel, and shorter residual queues at the downstream intersection — which is why corridor-level benefits compound beyond raw minute savings.

Quick answers

Sources: Cnn, Reddit, Reddit, Reddit, arXiv

Also worth reading: How smart cities work and the technology shaping the future of urban planning: How smart cities work and · How sand technology is redefining urban innovation and the future of smart cities: How sand technology is redefining · How smart city technology transforms the future of urban development: How smart city technology transforms