How Traffic Lights Actually Work

A traffic signal looks like three coloured bulbs on a pole. The engineering underneath is a real-time control system coordinating dozens of independent sensors, managing competing vehicle and pedestrian demands, synchronising with signals hundreds of metres away, and failing safely under hardware faults. Across Europe, adaptive traffic control systems adjust thousands of signal timings every few seconds in response to measured traffic. The difference between a well-tuned signal network and a poorly configured one shows up in commute times, fuel consumption, and ambulance response times.

This article covers how traffic signals work from the detector hardware and controller electronics through fixed-time and actuated control, adaptive systems deployed across European cities, signal coordination for arterials, emergency vehicle preemption, LED optics, and the V2X communication layer that is now appearing on European roads.

What a Signal Controller Actually Is

A traffic signal controller is an embedded computer dedicated to sequencing the signal indications (red, amber, green) for all approaches to an intersection according to timing plans, detector inputs, and coordination commands from a central system. It is not a timer with a light attached.

In European installations, the hardware architecture has historically followed national standards. Germany uses controllers compliant with the Technische Lieferbedingungen für Streckenstationen (TLS) protocol family. The Netherlands uses CCOL (Coupled Control Of Traffic Lights). The United Kingdom uses SCOOT-compatible UTC (Urban Traffic Control) controllers. Across these implementations, the functional structure is similar: a central processing unit executing the phasing logic, a set of output drivers for the signal lamp circuits (now almost universally LED-based), a set of detector input ports, and a communications interface for central system connection.

The safety-critical component inside every controller is the conflict monitor: a hardwired logic circuit (not software) that prevents incompatible signal combinations from appearing simultaneously. Incompatible means that two movements with physical crossing paths would both receive green indications. If the control software generates a signal output that the conflict monitor determines is incompatible, the monitor overrides the output, forces all signals to flashing amber (or red), and generates a fault log. The conflict monitor operates independently of the main processor. It cannot be reprogrammed through the normal controller interface. Its sole function is to prevent collisions caused by software or configuration errors.

Phase Structure and Ring-and-Barrier Diagrams

A signal cycle is divided into phases. Each phase is a combination of movements (vehicle lanes, pedestrian crossings, cyclist signals) that are allowed to proceed simultaneously without conflict. Two left-turn lanes at opposite ends of an intersection can share a phase; a left turn and an opposing through movement typically cannot (though with separate left-turn arrows and protected-permissive operation, they may under specific conditions).

The ring-and-barrier diagram, standard in traffic engineering, represents the phase structure graphically. Rings contain sequences of non-conflicting phases; barriers represent points where all rings must simultaneously transition. A simple four-leg intersection might have two rings with two phases each, and one barrier separating the north-south phases from the east-west phases.

The controller executes phases in sequence, spending a minimum and maximum time in each phase, extending based on detector input, and then transitioning through a clearance interval (amber plus all-red) before the next phase receives a green indication.

Inductive Loop Detectors

The most widely deployed vehicle detection technology in European signalised intersections is the inductive loop detector. A loop of wire is saw-cut into the asphalt surface of a lane and connected to a detector unit mounted in the controller cabinet. The detector drives an oscillating current through the loop at a frequency typically between 10 kHz and 100 kHz, forming an LC oscillator circuit with the loop's inductance.

When a vehicle passes over the loop, the steel and iron in the vehicle's body and chassis couples magnetically with the loop, reducing the loop's effective inductance. The inductance change is small, typically 0.01 percent to 1 percent of the unloaded value, but the detector's frequency counter is sensitive enough to measure it. A presence detection loop (typically 2 m × 2 m or 1.8 m × 2 m) registers a vehicle as present whenever the inductance is depressed below a threshold, and clears when the inductance recovers. A count loop (2 m × 6 m) is tuned to detect the leading and trailing edges of vehicles to count passages and estimate vehicle length.

Speed measurement requires two closely spaced loops (typically 3 to 6 m apart). A vehicle detected at loop 1 at time t₁ and at loop 2 at time t₂ was travelling at approximately d / (t₂ - t₁), where d is the loop separation. Vehicle length can be estimated from the time the vehicle occupies a single loop at known speed.

Loop detectors have limitations. Saw-cutting asphalt creates a water ingress path; freeze-thaw cycles in northern European climates (Scandinavia, Germany, Poland) degrade the sealant and eventually crack the wire, causing detector failure. Heavy traffic and road resurfacing work damage loops. Replacement requires cutting new slots, which is disruptive and expensive.

Video detection, using cameras mounted on signal poles or gantries, has displaced loops in many new European installations. A video detector processes frames from a standard or infrared camera using background subtraction and blob tracking to define virtual detection zones. The detector outputs the same presence and count signals as a physical loop without any in-pavement installation. The limitation is sensitivity to lighting conditions, camera contamination, and occlusion between vehicles.

Radar detectors (both Doppler and FMCW types) are increasingly common, particularly for detecting cyclists and motorcycles that inductive loops sometimes miss due to low ferrous content. FMCW (Frequency-Modulated Continuous Wave) radar can measure range and velocity simultaneously, classifying detected objects by size and movement pattern.

Fixed-Time Control and Webster's Formula

The simplest signal control strategy is fixed-time control: the controller runs a pre-programmed timing plan repeatedly without reference to detector inputs. The timing plan defines cycle length, phase sequence, green split for each phase, and offset relative to a time reference.

Fixed-time plans are designed using historical traffic flow data, typically from tube counters or multi-hour video surveys. The fundamental design tool is Webster's formula for optimal cycle length:

C = (1.5L + 5) / (1 - Y)

Where:

• C = optimal cycle length in seconds
• L = total lost time per cycle in seconds (sum of start-up lost time and clearance time for each phase)
• Y = sum of the critical lane flow ratios (one per phase)

The critical lane flow ratio for a phase is the highest volume-to-saturation-flow ratio among the lanes active in that phase:

y_i = q_i / s_i

Where q_i is the arrival flow in vehicles per hour for the critical lane and s_i is the saturation flow rate (the flow the lane can discharge when continuously green, typically 1,800 to 1,900 passenger car units per hour per lane in European conditions).

Once cycle length is established, the green time for each phase is allocated proportionally to its critical lane flow ratio:

g_i = (C - L) × y_i / Y

This formula allocates green time in proportion to demand, which maximises total intersection throughput. The minimum green time for each phase is constrained by pedestrian crossing requirements regardless of vehicle demand: pedestrians need time to perceive the signal change, step off the kerb, and cross, which EN 13201 (road lighting) and national guidance documents translate into minimum walk and clearance times based on crossing distance and pedestrian walking speed.

Fixed-time plans are designed for average conditions. A plan optimised for 8:30 AM Thursday conditions may perform poorly on a day with a football match nearby, a market, or a broken-down lorry blocking a lane. Most fixed-time-controlled intersections have three to five timing plans and switch between them based on time of day and day of week, but the plan selection is schedule-based rather than condition-based.

Vehicle-Actuated Control

Vehicle-actuated control extends fixed-time control by using detector inputs to modify green times within a cycle. The controller still follows a fixed phase sequence, but the duration of each phase depends on real-time detector readings.

The key concepts are minimum green, maximum green, unit extension, and gap-out.

Minimum green is the shortest time a phase can display green before it can be terminated. It accounts for vehicles that were already in the intersection or very close to the stop line when the phase started.

Maximum green is the longest time a phase can display green regardless of demand, preventing a single high-demand approach from monopolising the cycle and starving other movements.

Unit extension (also called vehicle extension or gap time) is the additional green time added each time a vehicle is detected by a detector typically located 40 to 60 m upstream of the stop line. If a vehicle is detected while the phase is green, the phase extends by the unit extension time (typically 2 to 4 seconds). This allows continuous traffic flow to hold a green phase active.

Gap-out occurs when no vehicle has been detected for a time equal to the gap threshold. The controller interprets this as the end of the vehicle queue and terminates the phase (subject to minimum green having elapsed and maximum green not having been reached). The gap threshold is typically set to the time for a vehicle at the detector location to reach the stop line, ensuring the last detected vehicle clears the junction before amber.

Semi-actuated control applies actuation only on minor approaches: the major arterial runs on a fixed background cycle, while side-street phases are only served when vehicles or pedestrians are detected. This suits intersections where one approach carries the dominant flow and interrupting it should be minimised.

Fully actuated control applies extension and gap-out to all phases. At low-traffic times (late evening, weekends), fully actuated intersections respond almost entirely to real-time demand, often skipping phases entirely when no demand exists on a movement.

Adaptive Systems: SCOOT, UTOPIA, and SCATS

Vehicle-actuated control optimises the timing of individual intersections responding to local detector inputs. Adaptive traffic control systems coordinate networks of intersections, adjusting timing plans across an entire urban area in response to real measured traffic. The distinction matters because urban traffic is interconnected: a queue on one link feeds back to the upstream intersection, and poor coordination between adjacent signals can cause queues to block junctions.

SCOOT

SCOOT (Split, Cycle, and Offset Optimisation Technique) was developed by TRL (Transport Research Laboratory) in the United Kingdom in the 1970s and 1980s and is now deployed in over 200 cities worldwide, including London, Edinburgh, Southampton, Rotterdam, and Gothenburg.

SCOOT uses inductive loop detectors located upstream of each stop line (at a position where the detector can observe vehicles arriving at the back of the queue rather than at the stop line itself). From these detectors, SCOOT builds a cyclic flow profile for each link: a model of how traffic is distributed within the signal cycle, representing the density of vehicles as a function of their position in the cycle.

From the flow profiles, SCOOT constructs a queue model for each link and predicts how signal timing changes will affect queue growth and dissipation. The optimiser then makes three types of small incremental adjustments every few seconds:

• Split optimisation: adjusting the green time split between phases at each intersection by plus or minus a few seconds to balance demand.
• Cycle length optimisation: adjusting the common cycle length for a group of coordinated intersections by plus or minus a few seconds to better match overall demand.
• Offset optimisation: adjusting the timing offset between adjacent intersections to improve platoon progression, allowing groups of vehicles released by one green to arrive at the next signal as it turns green.

SCOOT's adjustments are deliberately small and gradual. A change of more than a few seconds per cycle would cause vehicles already in transit between intersections to experience a disrupted progression. The small-step approach trades fast response speed for stability and comfort.

London's SCOOT network, managed by Transport for London, covers over 4,400 signalised intersections. Adjustments propagate across the network continuously, responding to incidents, special events, and daily demand variations without operator intervention.

UTOPIA and SPOT

UTOPIA (Urban Traffic OPtimisation by Integrated Automation) was developed in Italy in the 1990s by a consortium including SWARCO, Mizar Automazione, and research groups from the Politecnico di Torino. It has a hierarchical structure: SPOT (Single intersection tool for traffic optimisation) runs at each intersection and feeds data to the UTOPIA area controller.

SPOT performs real-time cycle-by-cycle optimisation at each intersection using local detector data, computing green times that minimise a cost function including delay and number of stops. UTOPIA coordinates the SPOT controllers across an area, providing offset targets and cycle length guidance to achieve network-level coordination.

UTOPIA is deployed in Turin (where it originated), Athens, Copenhagen, Bilbao, and several other European cities. The Athens installation, implemented in stages since the early 2000s, covers several hundred intersections in the central urban area and is managed by the Athens Traffic Management Centre.

SCATS

SCATS (Sydney Coordinated Adaptive Traffic System), developed in Australia by the Roads and Maritime Services of New South Wales, operates differently from SCOOT. Where SCOOT continuously adjusts in small increments, SCATS selects from a library of pre-computed timing plans based on current detector readings. The system measures the degree of saturation on each link (the ratio of vehicle count to capacity) and selects the plan that best fits the measured demand pattern.

SCATS is deployed in Dublin, Limassol, and several Greek provincial cities. Its plan-library approach makes it more predictable than fully continuous optimisation but less responsive to rapid demand changes.

Signal Coordination and the Green Wave

When a driver travels along an arterial road in Athens or along a boulevard in Copenhagen and finds the signals turning green just as they arrive, they are experiencing the result of offset optimisation: the systematic timing of signals so that a platoon of vehicles released from one green phase arrives at the next intersection as its signal turns green.

The fundamental calculation for a simple one-direction arterial is:

offset = distance / design speed

Where distance is the separation between stop lines and design speed is the speed at which the progression is designed (typically close to the posted speed limit). A signal 300 m downstream from a reference signal, with a design speed of 50 km/h (13.9 m/s), requires an offset of approximately 21.6 seconds: the downstream signal should turn green 21.6 seconds after the upstream signal.

The usable bandwidth of a green wave is the fraction of the cycle during which a vehicle travelling at the design speed can pass through all signals without stopping. In a well-coordinated one-way arterial, bandwidths of 50 to 60 percent of cycle length are achievable. For a 90-second cycle, this means a vehicle has 45 to 54 seconds in which to cross the first stop line and travel through the entire coordinated section.

Two-way coordination is harder. Vehicles travelling in both directions simultaneously through the same set of signals cannot both achieve maximum bandwidth because the optimal offsets for each direction conflict. The MAXBAND algorithm, developed at MIT in the 1970s, finds the offset values that maximise the minimum bandwidth across both directions simultaneously. For a regular grid with uniform block lengths and similar demand in both directions, MAXBAND typically produces bandwidths of 30 to 45 percent per direction, significantly less than one-way but still far better than uncoordinated signals.

Coordination requires all signals in the group to operate with the same cycle length (or integer multiples of a common base cycle). When SCOOT or UTOPIA adjusts cycle lengths, it does so for groups simultaneously to maintain synchronisation.

Emergency Vehicle Preemption

When an ambulance, fire engine, or police vehicle approaches a signalised intersection under emergency conditions, signal preemption interrupts the normal cycle and serves the approaching vehicle's phase as rapidly as possible.

The most widely deployed preemption detection technology in Europe is optical: an infrared strobe emitter on the emergency vehicle, pulsing at a specific frequency (14 Hz is the GTT Opticom standard, widely used in Germany, the Netherlands, and Scandinavia), is detected by a photodetector mounted on the signal pole. The controller verifies the correct pulse frequency to reject false triggers from other light sources.

Upon confirmed preemption request, the controller executes a preemption sequence:

1. Dwell: the active phase continues until it has been displayed long enough for traffic to clear (typically a few seconds of additional green, or the remainder of the minimum green if it has not elapsed).
2. Transition: all active phases are terminated in the normal way (amber, then all-red clearance interval).
3. Reservice: conflicting movements are cleared, typically by holding all-red for a fixed time to allow the junction to empty.
4. Preempt phase: the signal group on the emergency vehicle's approach receives green. All other movements receive red.
5. Return: after the preempt phase has been served for a minimum time and the detector clears (vehicle has passed), the controller returns to normal operation, typically resuming mid-cycle rather than starting from the beginning.

GPS-based and V2X-based preemption (see the section on C-V2X below) allow the approaching vehicle's trajectory to be known precisely enough to begin the transition sequence earlier, reducing both the vehicle's delay and the disruption to cross-traffic. A V2X-equipped ambulance approaching an intersection in Amsterdam or Helsinki can send a preemption request 300 to 400 m from the stop line, giving the controller approximately 20 to 25 seconds at 50 km/h to begin transitioning before the vehicle arrives.

Public transit priority (bus or tram priority) uses the same detection and preemption infrastructure but with different parameters. Rather than full preemption (which gives the transit vehicle unconditional green), transit priority extends an active green or recalls a green phase if the approaching vehicle is running behind schedule. On-schedule vehicles may receive no priority at all, limiting unnecessary disruption to cross-traffic.

LED Technology and Optical Design

Until the mid-2000s, traffic signal lanterns used incandescent bulbs: 100 W for pedestrian signals, 150 W for vehicle signals. A typical intersection with twelve vehicle signal heads and four pedestrian signal heads consumed approximately 2,000 W continuously from its signal lamps alone. Across a city of 1,000 signalised intersections, lamp power is 2 MW.

LED traffic signal lanterns have reduced this by approximately 85 to 90 percent. A modern LED vehicle signal head operates at 10 to 18 W. Pedestrian LEDs run at 5 to 10 W. The same intersection with LED lanterns uses around 250 W, and a network of 1,000 intersections uses approximately 250 kW. The payback period for LED retrofits in most European cities, accounting for lamp replacement labour savings, was 2 to 4 years, making the transition economically obvious regardless of environmental considerations.

The optical design of an LED signal head is more demanding than simply replacing the incandescent bulb with an LED array. Traffic signals must be clearly visible across a range of conditions: direct low-angle sunlight (the "phantom" problem where sunlight enters the lens and makes a red head appear to glow), bright diffuse overcast light, night, rain, and fog.

The phantom problem is particularly relevant to LED signals because the LED array has low enough luminance to be washed out by sunlight in the lens, making an unlit head look green from a distance. Solutions include improved visor geometry (deep visors that block low-angle sunlight) and anti-phantom optical filters. European standard EN 12368 (traffic signal lanterns) specifies the luminous intensity requirements in candela across the viewing angles and the chromaticity coordinates (defined within the CIE 1931 colour space) for each colour:

• Red: chromaticity within a specified polygon in the CIE diagram, peak wavelength approximately 630 nm.
• Amber: CIE coordinates within a defined polygon, peak wavelength approximately 590 nm.
• Green: CIE coordinates within a defined polygon, peak wavelength approximately 520 to 530 nm.

These chromaticity specifications ensure that the colours remain distinguishable under a range of viewing conditions and under sunlight. LED manufacturers must characterise their devices against these specifications, and EN 12368 testing verifies that the complete lantern (LED array, optical system, and housing) meets the requirements at end-of-life luminous intensity as well as initial.

LED failure mode differs from incandescent failure. An incandescent bulb fails suddenly and completely: one moment it is on, the next it is dark. An LED array typically degrades gradually: individual LEDs in the array fail one at a time, and the head appears at first as having dark spots, eventually dimming below visibility thresholds. A dark spot in a red head might be interpreted by a driver as a flashing amber, and a fully failed green that appears dark is obviously dangerous. Maintenance regimes for LED signals must include periodic luminance measurements rather than simply waiting for complete failure. Some controller systems monitor lamp circuit current and flag anomalies suggesting partial LED failure.

V2X Communication: ETSI ITS-G5 and C-V2X

Vehicle-to-infrastructure (V2I) communication allows signals to broadcast their current state and upcoming phase timing directly to equipped vehicles, and allows vehicles to send preemption requests or provide trajectory data to the controller. In Europe, this is being deployed under two competing but related standards.

ETSI ITS-G5 uses the 5.9 GHz band designated by European regulators for Intelligent Transport Systems. The physical and MAC layers follow IEEE 802.11p (also called DSRC, Dedicated Short-Range Communications), adapted for vehicular environments where communication time may be very short (a vehicle at 90 km/h is within 100 m range for approximately 4 seconds). The communication range is typically 150 to 400 m depending on antenna placement and environment.

Two standardised message types are relevant to traffic signals:

SPAT (Signal Phase and Timing): the controller broadcasts the current phase state (green, amber, red) and the remaining time in that state for each signal group, along with the expected timing of upcoming phases. An equipped vehicle receiving SPAT knows exactly when the signal will change and can compute an advisory speed to arrive at the signal at the start of the next green phase.

MAP (Map Data): describes the physical geometry of the intersection, relating signal groups to specific lanes and approaches. A vehicle receiving MAP can determine which SPAT messages apply to its intended path through the intersection.

GLOSA (Green Light Optimal Speed Advisory) is the application layer service that uses SPAT and MAP together. The vehicle's navigation system receives the current signal state and timing, computes the distance to the stop line, and advises the driver (or autonomous driving system) of a speed that will result in passing through on green without needing to stop. GLOSA implementations in European field trials (Germany, Netherlands, France under the C-Roads platform) have demonstrated fuel savings of 5 to 15 percent on equipped vehicles approaching coordinated arterial sequences, because smooth deceleration and re-acceleration is more efficient than stop-start driving.

C-V2X (Cellular Vehicle-to-Everything), standardised by 3GPP as LTE-V2X (Release 14) and extended to 5G NR-V2X (Release 16), is an alternative physical layer using LTE or 5G radio technology for direct vehicle-to-vehicle and vehicle-to-infrastructure communication. C-V2X in direct mode (sidelink operation, also called PC5 interface) does not require a cellular network connection and uses the same 5.9 GHz band as ETSI ITS-G5. The message formats (SPAT, MAP, and others defined by ETSI TC ITS standards) are the same across both physical layers.

Europe has not yet mandated a single standard, and the C-Roads platform (a European Commission co-funded initiative involving road operators across Austria, Belgium, Czech Republic, Denmark, France, Germany, Hungary, Netherlands, Spain, and others) is deploying both technologies in parallel to gather operational experience. As of 2024, over 1,000 European intersections have been equipped with ITS-G5 or C-V2X roadside units, with major deployments on A2 and A10 motorway access routes in the Netherlands, the A9 in Bavaria, and the périphérique access points in the Île-de-France region.

Failure Modes and Fallback Behaviour

A traffic signal system must fail safely. Safe does not mean convenient: safe means the failure mode does not cause collisions.

Controller Power Loss

If the controller loses mains power, all signal heads go dark: no green, no amber, no red. This is the most dangerous failure mode because drivers approaching a dark signal may not recognise it as a signalised intersection at all, particularly at night or in unfamiliar areas. Most European controllers have a battery backup or UPS capable of sustaining operation for 1 to 8 hours, sufficient for brief power interruptions. When the battery is exhausted or if there is no UPS, the intersection should be treated as an uncontrolled junction, with right-of-way rules applying.

Flash Mode

When a controller detects a hardware fault it cannot recover from, or when it is commanded into a maintenance state, it enters flash mode: all signal heads display a repeating amber flash (or red flash on the cross streets, in some national implementations). Flashing amber signals all approaches to treat the intersection as a yield situation. This is safer than darkness because drivers are clearly informed of a failure state. The conflict monitor forces flash mode if it detects an incompatible signal output.

Detector Failure

When a loop detector fails (open circuit from a broken wire, or short circuit from moisture ingress), the controller loses information about vehicle demand on that approach. The fallback behaviour depends on controller configuration. Options include:

• Treating the failed detector as continuously occupied (the approach always has demand): this prevents the gap-out mechanism from terminating the phase early, but may cause the phase to always extend to maximum green even at low traffic, adding delay on other approaches.
• Treating the failed detector as continuously empty: the phase gaps out immediately after minimum green, which may be insufficient for the actual queue.
• Falling back to a fixed timing plan for the affected phase: the simplest and most common fallback.

Communication Loss

If a controller loses its communication link to the SCOOT or UTOPIA central system, it falls back to a locally stored timing plan. The local plan is typically the plan that was active at the time of communication loss, or a default plan for the current time of day. The intersection continues to operate correctly in isolation; it simply loses the network-level coordination and adaptive optimisation benefits. Operators at the Traffic Management Centre are alerted, and engineers are dispatched to restore the communication link.

From Sensor to Signal: The Complete Picture

A vehicle approaching a signalised intersection in central Copenhagen, travelling on a green phase that is about to gap out, triggers the following sequence: the inductive loop 50 m from the stop line detects the vehicle's inductance signature, the detector unit signals presence to the controller's input card, the control processor registers the extension request, the current phase timer is reset by the unit extension value, the UTOPIA SPOT module increments the flow count on that link, and the network optimiser adjusts its model of current demand. If the vehicle is an ambulance with an Opticom strobe active, the preemption detector fires simultaneously, overrides the normal sequence, and initiates the preemption transition. If the vehicle is a connected car with ITS-G5 capability, it has already received the SPAT message 300 m back, computed a GLOSA advisory, and adjusted its approach speed so that it would have arrived at the start of the green phase without needing the extension at all.

None of these systems operates independently. The detector, the controller, the coordination network, the preemption system, and the V2X broadcast are all layers on top of the same physical signal head. The traffic light on the street corner that a pedestrian treats as a simple red or green is the visible output of a control system considerably more complex than it appears.

Pedestrian Signal Logic

Vehicle phases and pedestrian phases at the same intersection must be carefully coordinated, and the rules for pedestrian signals are more conservative than those for vehicles because pedestrians are more vulnerable and move more slowly.

A pedestrian crossing phase has three intervals: walk, pedestrian clearance (flashing don't walk or countdown), and all-red or vehicle clearance. The walk interval begins when the pedestrian green signal activates. Its minimum duration is based on reaction time plus the time for the fastest expected pedestrian to perceive the signal change and step off the kerb: this is typically 7 seconds at a minimum regardless of crossing width. The pedestrian clearance interval must provide enough time for a pedestrian who entered the crossing at the end of the walk interval to reach the opposite kerb (or a pedestrian refuge island). European standards and national guidance use a pedestrian walking speed of 1.2 m/s for adults in typical daytime conditions. Many authorities apply a lower design speed (1.0 m/s or even 0.8 m/s) where elderly pedestrians are common, such as near hospitals or care facilities.

For a 20 m crossing at 1.2 m/s, the clearance time must be at least 16.7 seconds. Adding the walk minimum of 7 seconds and a conservative buffer, the total pedestrian phase minimum is typically 25 to 30 seconds. This is a significant allocation of cycle time and is why pedestrian phases at wide crossings on long cycle signals can consume 30 to 40 percent of the total cycle.

Pedestrian detection is used to avoid serving pedestrian phases when no pedestrians are present, which is important at night or on quiet side streets where a mandatory pedestrian phase every cycle delays vehicles unnecessarily. A push button records pedestrian demand; the controller serves the pedestrian phase when demand is registered (similar to vehicle actuation) rather than on every cycle. Some installations use passive infrared (PIR) detectors or video cameras to detect pedestrians waiting at the crossing, allowing the controller to extend the walk interval if pedestrians are still crossing when the clearance interval is due to begin.

Accessible pedestrian signals add audible and tactile cues for visually impaired users. The audible signal is typically a slow tick during the don't walk phase and a fast tick during the walk phase, or a specific sound (a cuckoo or other call) per approach direction. EN 14380 (tactile indicators at pedestrian crossings) and national standards govern the placement of tactile paving (blister paving in the UK, striped warning paving in Germany) at the crossing point and the design of the rotating cone or vibrating button at the push-button post that supplements the audible signal for deaf-blind users.

Countdown timers at pedestrian crossings, now standard in many European cities, display the remaining walk or clearance time in seconds. They improve pedestrian compliance with signals because pedestrians can judge whether they have time to cross rather than guessing. Studies from cities including Oslo and Lisbon show reduced red-light violations at installations with countdown timers, particularly during the clearance interval when pedestrians previously would often continue entering a crossing.

Saturation Flow and Intersection Capacity

Understanding how much traffic a signalised intersection can handle requires the concept of saturation flow: the maximum rate at which vehicles can discharge from a stopped queue at a stop line when continuously given a green signal.

Saturation flow is measured empirically by observing a queue discharging. The first few vehicles in the queue discharge more slowly than steady state (start-up lost time, typically 2 seconds per phase, accounting for reaction time and the mechanical process of vehicles beginning to move). Once the queue is flowing freely, the discharge rate stabilises at the saturation flow rate. This rate depends on lane width, vehicle mix, proximity to kerbs and turns, and gradient:

• A standard 3.65 m lane in European conditions: approximately 1,800 to 1,900 passenger car unit equivalents per hour (pcu/h).
• A narrow 3.0 m lane: approximately 1,500 to 1,600 pcu/h.
• A lane with significant heavy vehicle traffic: lower pcu/h because heavy vehicles occupy more time in the stop line gap.
• A turning lane (vehicles must slow for the turn): 1,400 to 1,600 pcu/h depending on turn radius.

Passenger car unit equivalents (pcu) normalise different vehicle types to a common scale. A passenger car is 1 pcu. A heavy goods vehicle on a through movement is typically 2.0 to 2.5 pcu. A cyclist at an advanced stop line is 0.2 to 0.5 pcu. The pcu factor for each vehicle type is derived empirically from the additional time that vehicle type consumes at the stop line relative to a passenger car.

The capacity of a lane during a cycle is:

Q = s × (g / C)

Where Q is the lane capacity in pcu/h, s is the saturation flow rate, g is the effective green time for that phase, and C is the cycle length. The ratio g/C is the green time fraction, sometimes called the degree of green. For a 90-second cycle with 40 seconds of effective green for a through phase, the green fraction is 0.44. A lane with saturation flow of 1,800 pcu/h achieves a capacity of 1,800 × 0.44 = 792 pcu/h.

Degree of saturation (x) is the ratio of demand to capacity:

x = q / Q = q × C / (s × g)

Where q is the arrival flow in pcu/h. When x approaches 1.0, the approach is near capacity: queues grow each cycle and do not fully discharge before the next red phase. When x exceeds 1.0, the approach is over-saturated: queues grow indefinitely until demand falls below capacity. Signal design targets keep x below approximately 0.9 on each critical lane to maintain stable operation and provide a buffer for demand variation.

Webster's formula for optimal cycle length minimises total vehicle delay at the intersection, which it achieves by keeping all approaches operating at similar degrees of saturation below 0.9. Cycle lengths that are too short waste a large fraction of the cycle in start-up lost time (low effective green). Cycle lengths that are too long give excessive green to low-demand phases while others queue unnecessarily. The optimal cycle balances these competing losses.

Measuring and Auditing Signal Performance

A signal controller running an adaptive system generates continuous measurements of detector activity, queue lengths, and phase service times. A well-resourced traffic management centre uses this data to audit system performance, identify misbehaving detectors, and tune timing parameters.

The primary performance metric for a signalised network is average control delay per vehicle: the extra time a vehicle spends at the intersection compared to what it would spend if the signal were not present and the intersection were empty. Control delay has two components: stopped delay (time sitting at a red signal) and deceleration-acceleration delay (the time lost in braking to a stop and re-accelerating to speed, which is typically 3 to 5 seconds per stop even at a short red).

HCM (Highway Capacity Manual) methodologies and the European Saturation Flow Manual provide formulae for estimating delay from timing parameters and flow measurements, but the most accurate measurement is from floating car surveys: an instrumented vehicle (or smartphone app) drives through the network repeatedly at different times and records travel time on each link. Comparing observed travel times to free-flow travel times yields the delay attributable to signals.

Probe vehicle data from connected cars and GPS-equipped mobile phones (processed anonymously and in aggregate) has become an important supplement to fixed detector data in European cities. Amsterdam's traffic management centre uses anonymised GPS traces from navigation apps combined with SCOOT detector data to identify unexpected queue spills and adjust signal timing in near-real-time. Athens has piloted a similar data fusion approach for sections of the Kifissias Avenue corridor.

Stop rate (the fraction of vehicles that have to stop at least once) is a secondary metric that correlates with fuel consumption and emissions. A GLOSA system that reduces stop rate from 70 percent to 40 percent on an arterial delivers a measurable reduction in NOx and particulate emissions, which is increasingly relevant for EU cities under air quality compliance pressure.

Detector health monitoring is an underappreciated part of adaptive system performance. A failed detector that reports permanently empty causes the associated phase to always gap out at minimum green, potentially under-serving an approach that is actually busy. A failed detector that reports permanently occupied causes the phase to always extend to maximum green, wasting capacity on a possibly empty approach. SCOOT and UTOPIA both include detector monitoring routines that flag suspicious activity (a detector that never clears, or a detector that records zero vehicles over a period when adjacent detectors show normal activity). Automated alerts to maintenance staff are standard in modern deployments.

Signal timing review is a periodic process: even adaptive systems benefit from reviewing the base parameters (saturation flow estimates, lost time values, minimum and maximum greens) annually and updating them as land use and vehicle mix change. New residential developments, large retail openings, or changes to public transit routes near intersections alter the demand pattern in ways that adaptive algorithms optimise around, but which may benefit from explicit timing parameter updates to reflect the new underlying conditions.

Tram and Cycle Signals

Traffic signals in European cities frequently need to accommodate tram lines and dedicated cycling infrastructure, both of which operate under different rules than motor vehicles and require dedicated signal hardware.

Trams present a specific challenge because they share road space with other traffic but cannot deviate from their track and cannot stop as quickly as a car. A tram approaching a signalised intersection needs priority treatment: either a dedicated tram phase that gives the tram exclusive access to the junction, or a transit signal priority call that extends the active green for the tram's direction if the tram is present. Tram priority is nearly universal in cities with significant tram networks: Amsterdam, Brussels, Vienna, Prague, Lisbon, and Athens all apply priority logic for their tram systems.

Tram detection uses transponders in the track (inductive loops tuned to the tram's resonance frequency, distinct from car loops) or GPS-based automatic vehicle location (AVL) systems that report the tram's position and predicted arrival time at each controlled intersection. The AVL approach allows the controller to begin the priority call earlier than a detector-based approach because the tram's position is known 200 to 400 m before it reaches the stop line.

Tram signals use dedicated signal heads displaying either the standard tri-colour for trams (which sometimes differ from road signals in the amber-skip sequences used) or the European tram signal, an F-shaped or bar-type display used in Germany and Switzerland to give tram drivers route and speed information independently of the road signals. The tram signal is not visible to car drivers and therefore does not create ambiguity about which vehicle is being directed.

Cycle signals address the different speed, vulnerability, and road position of cyclists. An advanced stop line (cycle box) allows cyclists to position themselves ahead of waiting cars, improving their visibility to drivers and giving them a head start when the signal turns green. Some cities (Amsterdam, Copenhagen, Helsinki) supplement cycle boxes with a small early-release green for cyclists: the cyclist signal turns green 3 to 5 seconds before the adjacent vehicle phase, allowing cyclists to clear the stop line before cars begin moving. This reduces conflict between turning vehicles and straight-cycling cyclists.

Cycle detection at signalised junctions uses video cameras or radar rather than inductive loops, since bicycles have minimal metallic content and are frequently missed by standard vehicle loop detectors. Video detection can identify a cyclist waiting at an advanced stop line and flag demand to the controller, ensuring the phase is served even if no other vehicles are present. Without dedicated cycle detection, a cyclist at an actuated signal with no other vehicles may find the phase never called, forcing a red-light run or a long wait for the next cycle.

Some Dutch and Danish cities have introduced dedicated cycle signal phases that run in parallel with pedestrian signals during the all-red vehicle clearance interval, allowing cyclists moving along a separated cycle path to proceed while all vehicle movements are stopped. This approach is only possible where the cycling infrastructure is fully separated from vehicle lanes, but where it applies it significantly increases cycling throughput without increasing vehicle delay.

The conflict between cycle paths and turning vehicles (right-turning vehicles crossing the path of straight-cycling cyclists, the opposite in left-hand-traffic countries) is one of the most complex signal design problems in European cities. Protected intersection designs, originating in the Netherlands and now adopted in Copenhagen, Berlin, and London, use a curved vehicle turn path that slows right-turning vehicles to 10 to 15 km/h and a separate cycle phase that prevents vehicles from entering the turn during the cycle green. The signal geometry for these intersections requires additional phases and often longer cycles, but the safety benefit from eliminating the turning-conflict is substantial: protected intersections show significantly reduced cyclist casualty rates compared to conventional junctions in Dutch data from the period 2010 to 2020.

Physical Infrastructure: Poles, Cabinets, and Power

The hardware visible on the street (signal heads, poles, push button posts) is the tip of a substantial buried infrastructure that most road users never consider.

Signal poles in European cities are typically 76 mm or 102 mm diameter tubular steel, hot-dip galvanised, with a base plate bolted to a buried concrete foundation. The height is typically 3.0 to 4.5 m for standard vehicle signals, chosen to place the signal head at a height clearly visible to drivers of cars and lorries from the required approach distance (typically 60 to 100 m for a 50 km/h road). Poles at complex intersections often carry cantilever arms to position signal heads over the lane they control rather than at the roadside, improving visibility and reducing ambiguity about which signals apply to which lane. Gantry structures spanning the full road width are used on multi-lane arterials where roadside poles cannot provide adequate sightlines.

The controller cabinet is a weatherproof enclosure, typically sheet steel or GRP (glass-reinforced polymer), mounted on a concrete plinth beside the junction. It houses the signal controller, the conflict monitor, fuse and isolator panels, the telecommunications interface (fibre, ADSL, or 4G/LTE modem), and in many installations a UPS battery. Cabinet size has grown as the number of signal groups, detector inputs, and communication functions has increased: a modern controller for a complex intersection with twelve signal groups, eight detector inputs, pedestrian push buttons, preemption detection, and V2X hardware may require a cabinet of 800 mm wide by 600 mm deep by 1,200 mm tall.

Power is supplied from the local distribution network at 230 V AC single-phase, with a dedicated supply from the distribution board of the nearest substation. In areas with frequent power interruptions (older city centres in southern European countries, including parts of Greece and Portugal), a UPS rated for 4 to 8 hours of full operation ensures continued signal function during brief outages. The total power consumption of a modern LED-equipped signalised junction including the controller, communications equipment, and LED signal heads is typically 500 to 800 W: a dramatic reduction from the 2,500 to 4,000 W consumed by an equivalent incandescent installation of the same scale.

Cable ducting runs beneath the footway and carriageway between each signal head, detector, push button post, and the controller cabinet. A typical four-leg intersection requires between 300 and 600 m of conduit, carrying individually screened cables for each signal lamp circuit, detector loop leads, push button circuits, and communications. The loops themselves are connected by twisted-pair lead-in cables from the saw-cut in the asphalt to a junction box on the footway, and thence to the cabinet. Getting this cabling right during installation is critical: poorly made connections in the detector lead-in are the most common cause of loop detector failures that do not involve physical damage to the loop wire itself.

Cable testing at commissioning involves verifying the loop inductance (typically 50 to 200 µH depending on loop size and depth of installation), the lead-in cable resistance, and the insulation resistance between the loop and earth. European installation standards require insulation resistance above 100 MΩ at 500 V DC for new installations. As loops age and the sealant around the saw-cuts degrades, insulation resistance falls, eventually causing false detection or complete failure. Preventive maintenance programs in northern European cities schedule loop resistance testing on a 3 to 5 year cycle and replace loops whose insulation resistance has fallen below a threshold before they fail completely.