Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

2.3.

4 Control

2.3.4.1 Fixed Timing

These are the simplest controls to consider. Traffic lights which are only capable of timed control

are obsolete having been replaced by VA systems. The light sequence and timing are both fixed

and as a result there is no need for traffic detection. To be effective, the controller needs to be

tuned to the actual roadway in terms of cycle time and split, giving priority to the main route if

this exists. As traffic trends change over time, the tuning will become dated and the junction will

operate below the theoretical optimal rate.

A major failing of timed systems is related to the inability to react to dynamic traffic flow volumes.

The timing necessary to provide adequate responsiveness while maximising throughput balanced

across at least two competing flow directions has to be calculated in advance. The calculation is

based on physical vehicle counts collected by observation over a long period of time and so fixed

timing does not respond to unanticipated “events”.

Fixed timing is very inefficient when the traffic flow rates are low. When flow rates are low, the

probability of a vehicle arriving is low but the random nature of traffic means the probability of

the vehicle arriving during a red phase is over 50% (assuming an equal split). There is an entirely

even probability of the vehicle arriving at any time during the red phase. Overall, a vehicle is more

likely to experience a delay than it is to avoid a delay. Flow losses due to signal changes can be

insignificant because of the low flow rates but an individual vehicle can be delayed by longer than

a full red cycle time if it arrives at the signal as the signal transitions to yellow; the vehicle will

then wait for the combined yellow, red and red/yellow periods. The low flow rate also implies that

the probability of a vehicle arriving at a conflicting direction is also very low therefore the initial
vehicle may wait without reason. The minimal number of vehicles is unlikely to require the entire

green time in any direction.

By contrast, fixed timing can be the most efficient method when junctions are over saturated as

the number of signal changes, and hence the flow losses due to signal changes, are minimised. The

implication of a congested junction is that traffic is waiting in all directions and because of this,

each green stage should be fully utilised.

Simple fixed time control is not in common use, possibly due to the cost/benefit advantages of

installing VA systems. Other more advanced systems can be forced to operate in fixed timing

mode where this is found to be advantageous. VA systems revert to an equivalent to fixed timing

during congestion.

Isolated, fixed timing junctions with 50% split which are busy but not congested are frequently

used as the benchmark to measure the improvement of proposed flow algorithms.

2.3.4.2 Vehicle Actuated

Vehicle Actuated (VA) systems resolve some of the problems of fixed timing by detecting traffic

approaching the signal and responding accordingly. Vehicles have the same probability of arriving

at a red indication as for timed systems but “calling for green” reduces the time delay experienced

when there is no conflicting flow. Original VA systems detected traffic at the stop line, changing

to green when required. Traffic lights change in the normal sequence and with a minimum green

period of typically 7 seconds. During the green phase, a vehicle detected on the approach causes

an extension of the green period. The extension may continue up to a maximum timer even if the

flow rate is relatively low, which deprives the conflicting traffic of flow priority and may seriously

impact junction efficiency. The detectors are set to “fail active” [29], meaning that a lost signal is

the same as a vehicle detected signal; failures may cause unnecessary green extension in a single
direction and a correspondingly unnecessary red in the conflicting direction. For flow efficiency

and to prevent instability, VA systems prioritize existing flows over new requests.

2.3.4.4 Adaptive Urban Traffic Controls (UTC) general

UTC systems are intended to resolve many of the limitations of earlier systems by combining

traffic signals into groups to improve traffic flow in an area or along a corridor. There are several

variations on the theme including UTOPIA (Urban traffic optimisation by integrated automation)

[27], SCATS (Sydney Coordinated Adaptive Traffic System) [33] and SCOOT (Split Cycle Offset

Optimisation Technique). Two of the most popular are systems are SCOOT and SCATS. A

common feature to these systems is the ability to modify the split and cycle times independently

i.e. the split percentage can be kept constant while the cycle length changes or the split percentage

can change within a fixed cycle length. In order to achieve this, the timing is modified after a cycle

completes and not during the current cycle as is the case with VA. The offset period between

consecutive lights is also variable. To prevent wide fluctuation in signal timings, the proposed

changes are confirmed over several cycles before being applied and timing changes are limited to

a few percent in any consecutive cycle meaning that several cycles may be required to apply the

entire change. There is therefore an element of prediction required to determine the response that

will be necessary a few cycles after the detected event. Adaptive controls are typically continually

self-tuning based on rolling historic data such as the flows in the previous hour, at this time on the

previous day, this day last week etc. The dynamic response of UTC may be impacted by filtering

and level of change limitation. The application of previously determined models and the reduced

dynamic response can minimise the ability of UTC to resolve unexpected and rapidly changing

traffic flows.
2.3.4.5 SCOOT

In simple terms, SCOOT counts upstream vehicles that are approaching the control point. By

recording the speed and number of vehicles, predictions are made regarding queue lengths and

when phase changes should be made. This system relies on traffic flowing at the detection point,

congestion is recognised when traffic is stationary at the detector [34] [35].

SCOOT is a complex, computer controlled Urban Traffic Control (UTC) system which is designed

to operate over multiple junctions (nodes) in multiple regions. SCOOT applies the results of

detected traffic to modelling system with the intention to estimate and predict queue lengths and

dispersals. Estimation allows the system to respond to the current situation whereas prediction

allows the system to prepare for future trends [36].

The basic operation of SCOOT is to detect traffic, model outcomes, compare to actual flow and

apply the most effective control strategy. The following information is gathered and collated from

a variety of sources and presentations including [35] [37] [34] [38]

SCOOT uses detection which is placed a distance upstream from the stop line, preferably

immediately after the previous traffic signal and beyond the distance that will be occupied by the

longest queue that is expected to form. The detected data is collected and converted to link profile

units (lpu) which combines aspects of flow and occupancy. The data over a period is compiled into

a cyclic flow profile (CFP) and patterns are extracted to determine, for example, platoon formation.

Metrics used to measure control effectiveness include Degree of Saturation (DegS) and

Performance Index (PI). Traffic controls within a region operate with the same cycle period to

maximise the flow. Region boundaries are predetermined and occur where the distance between

control signals is considered too great to allow effective co-ordination.


SCOOT has three optimisation systems one for split, one for cycle and one for offset, each of

which estimate the overall effect of small incremental changes in each attribute.

The split optimiser updates before the phase change of every controlled junction. The objective is

to determine if traffic flow would be most efficient with the current split time or with more or less

green time. Efficiency in this case is determined to be minimising the maximum degree of

saturation. The maximum amount of change is predefined and normally in the range 1 – 4 seconds.

The offset optimiser is run each cycle for every approach. Using modelled results and the CFP, the

optimiser attempts to improve the platoon (or green wave) c34apability of consecutive junctions

by adjusting the time differential of changes between upstream and downstream junctions. The

change is varied in 4 second increments.

The cycle optimiser is least frequently run, after every 2.5 or 5 minutes. The optimiser identifies a

single junction in each region as being critical and attempts to adjust the cycle time to maintain a

90% degree of saturation on each approach. Changes can either increase or decrease in 4, 8 or 16

second increments. Because the cycle time is consistent for all junctions within a region, any cycle

time changes affect all junctions in that region. The cycle length increases in direct proportion to

the traffic demand up to a maximum value.

Congestion management is achieved by gating (preventing further traffic from entering a region).

This has been described as being “…used to “relocate” traffic queues to links where they do not

affect critical approaches to junctions…” [39]

Other functions include prioritisation of specific vehicles, e.g. normal public transport, late (to

schedule) public transport. There is also facility for “emergency green waves”.

2.3.4.5.1 Limitations of SCOOT


There are many claims made regarding the benefits of SCOOT however it is clear that the system

is not perfect and does have limitations.

Reliance on solution models derived from earlier traffic flows is only viable while the current

flows conform to previously identified situations. By definition, an exceptional circumstance will

not have a model response and therefore a correct response cannot be applied. The rolling update

model system is intended to smooth out random events, removing wide variation; the response to

the exceptional circumstance will be diluted over a period of time until it is no longer available

even if the same event re-occurs.

SCOOT does not normally automatically skip phases or change sequence. Phase skipping is

available for prioritising busses when required however prioritising a single low frequency class

does not resolve congestion, it creates additional delays for the higher frequency classes i.e.

passenger cars. The prioritisation system can only provide benefit while the priority traffic is the

minority class. Priority systems may require additional lanes to be built or a restriction of the road

space available to other vehicle classes.

The system does not actively resolve existing congestion. The application of gating reduces

congestion within a specified region at the detriment of the surrounding approach regions. Gating

does not prevent or resolve congestion, simply moves congestion to a “more politically suitable”

location.

In common with all approach detection systems, the detection becomes ineffective when queueing

extends back to detector. The inference is an infinite queue and this cannot be efficiently resolved.

If an infinite queue is identified on more than one approach, a fair sequence cannot be identified

and there will be a reliance on manually collected turn data.


The upstream location of the detectors means that traffic sinks and/or sources that exist between

the detection and control point cannot be managed. This could be a particularly difficult problem

to resolve when there are temporary sinks and sources e.g. roadside retailers. The problem is

caused when a vehicle is counted entering the controlled route but then temporarily stops inside

the route and therefore does not pass during the modelled cycle; this is an inefficient use of the

upstream control sequence. When the vehicle continues the journey, it will reinsert within a cycle

that it has not been counted for which may result in queueing at a downstream signal.

The location of the detectors also means that individual lane detection is pointless; vehicles can

and will change lanes in the approach to the signals. The proportionate allocation of time at a traffic

signal requires previously collected turn data which may become stale over time or incorrect during

transient flow periods.

Cycle time optimisation for the entire region is based on the perceived best solution for a single

“critical” junction in the region. This might not be optimal for all junctions in the region especially

during the congestion build up and clearance phases. The optimisation attempts to prioritise a

single green wave path with the aim of resolving a predicted flow volume and direction.

A very important observation regarding the detection and update sequence is that SCOOT cannot

change the current cycle if congestion develops. Once the current response has been selected,

several cycles must pass before a further response can be applied.

2.3.4.6 SCATS

In similarly simple terms, SCATS detects vehicles at the stop line. Measuring flow rates during

the green phase and the waiting time at the stop line, it is possible to estimate the queue length

upstream from the control point. The SCATS detectors can also provide individual lane detection

allowing “phase skipping” when no traffic is detected [40].


SCATS calculates a best local solution based on current detection volumes. When adjacent signals

agree on the split and cycle periods, they undergo a marriage and synchronise the signals. At some

later stage, if adjacent signals disagree over timing the signals undergo a divorce. The marriages

and divorces system allows a degree of flexibility to determine and respond to the congestion

region boundary.

2.3.4.6.1 Limitations of SCATS

SCATS uses actual traffic measurements to determine a suitable response profile rather than using

a modelling approach. The detectors are closer to the lane ends and therefore per lane detection is

used. However SCATS determines the next suitable change over a period of cycles and is limited

by the change in split and cycle that is applied on each consecutive cycle. In common with SCOOT,

SCATS is therefore making a prediction regarding the solution that needs to be applied several

minutes after the actual detection. The solution will have transient peaks and troughs smoothed

out by the integrating action of multi-cycle averaging making the system less responsive to highly

dynamic changes

2.3.4.7 Intelligent Transport Systems

Future detection and reporting methods will include the use of in-vehicle sensors, V2V / V2I /

V2X communications and VANET (Vehicular Ad Hoc Networks) [41] [42] [43]. The exact

function of the detection and reporting is not fully defined but it is assumed that the detection will

in effect be a form of approach detection and there will be some form of third party arbitration and

signalling.

You might also like