the metropoliton toronto traffic control signal system

PROCEEDINGS OF THE IEEE, VOL. 56, NO. 4, APRIL 1968 577

The Metropolitan Toronto Traffic Control Signal System

J. T. HEWTON

Abstract-Faced with an ever-increasing economic loss due to tralXc cmgh ami accidents, the Metropolitan Toronto Aothorities investigated t h e ~ o f i m p r o v i n g c u n d i t i o n s ~ t h e t h e o f a g m e r a l - p o r p a r e digitrlcorlrtertoopthnizeinrealtimenndonliwtheoperntioooftr~c collbd FoUowing the successfal coocllsion of a d s c a l e pilot Icopct, appd was obtained for a completely integrated system designed to korpomte snne loo0 signak distributed in complex network formation -am area of some 240 square miles.

C o m e the ccmcept into reality required the d e s b pod installation of a eorpilerrble anmmt of specialized equipment, & pf vehicle detectors, togetk with the development and preparation of complex computer pro- grrap - both control and analysis rootioes. With the initinl stages of

meeed, sbrtisg nith the least sophisticated and gradually working up, while m y stdying actnal traffic coaditiols so that kvek of performance

After f o r years’ effort, some 500 out of 850 existing signak are d e r coolpta conlrol for an average of 15 hours every day, with the result that the tie rrq.ired to make any given polney has been reduced by an average of a b d 10 meat, even though the operational flexibility is stiU to some extent Liited. A@ from thk major improvement, sdkient other benefits both direct a d hdirect have resulted from system operation to prove the validity of tLe aighrl coocept, especially on economic grow&. A cuoservative extimate iiicptes that, for a total capital outlay of some S5 million, wastage at trkbbk to congestion and delay will be redneed by some $20 million pa Ye=-

this work more oc less cwnplete, testing of V a r i O o E control philosophies com-

eoalabefid.

1 INTRODUCTION

N THE TEN YEARS prior to 1957 the Metropolitan Toronto Region underwent a period of extremely rapid growth, during which population increased by over 38

percent and vehicle registrations by some 56 percent. Un- fortunately, the development of facilities for t r a c movement did not keep pace with the increasing demand. Thus congestion had become commonplace and of sufficient magnitude to make it obvious that something must be done to relieve the situation. Since all indications were that the growth rate would continue or even increase, it was even more obvious that any solution to the traflic problem must provide reasonably immediate relief, while including at least the initial steps toward a more efficient arrangement for urban transportation.

It was decided that the following plan provided a realistic solution which would be acceptable to the public, economically feasible, and capable of reasonably rapid implementation.

1) To provide for the future, mass rapid transit, such as subways, would be developed to serve the core area

Manuscript received September 18, 1967; revised December 11, 1967. The author was formerly Director of the Metropolitan Toronto Traffic

Control Centre, Toronto, Ontario, Canada. He is now President of J. T. Hewton and Associates, Toronto, Ontario, Canada.

and adequate low-cost parking facilities would be established at outlying points to encourage its use. At the same time, existing streets and highways would be improved and the construction of new facilities such as expressways undertaken. All of this would be accomplished as quickly as money could be made available.

2) To provide immediate relief, the traffic handling capability of the existing streets would be increased.through the development of efficient movement plans implemented by the operation of a traffic control signal system.

At that time, almost all of the traffic control signals in metropolitan Toronto operated on a pretimed basis, many with only a single timing dial. There was no effective system control, although in the downtown area simple interconnection provided for the coordination of a few individual groups in any one of three arrangements. It was obvious that a considerable expenditure would be necessary in order to achieve the desired results, and there was considerable doubt as to which, if any, of the many specialized systems available would be most satisfactory. On examination, they all suffered from defects as follows:

Specdized equipmeat-of the analog type specifically designed and built for a particular purpose would be necessary for both individual intersection and central control. Thus, any change in phasing arrangements would require the replacement of hardware and there would be no possibility of adopting or even trying new concepts in the theory or practice of traffic control. Though an individual intersection could operate on a fully traffic-responsive basis, the central control equipment could, at best, only use information from a few sampling detectors to select a predetermined operational plan. The preparation of the large number of operational plans necessary for a complex system would be a matter of compromise between many conflicting requirements, and therefore would tend to proceed on a trial-and-error basis with no assurance that optimum results had been achieved. The traffic movement information on which any operation plan must be based could not possibly be collected in a short time or even simultaneously throughout the area. The result, therefore, would not be optimum for presently existing conditions, but only for those which had existed at some considerable time in the past, if indeed they had ever existed at one and the same time.

578

In view of these limitations it was felt that this type of system operation, while no doubt providing some improvement over existing conditions, would be static, tending to force vehicular movement into preconceived patterns instead of dynamically adjusting to meet immediate needs. At this point in time a report was received from the Traffic Research Corporation which suggested that if a general- purpose digital computer were operated in real time and on line it could:

take in traffic information from a large number of vehicle detectors, determine the interval lengths required at each individual intersection, and optimize these for overall system efficiency considering existing conditions; determine the optimum time relationship, or offset, between individual intersections considering the existing traffic speed and direction of flow; directly control the individual signals to produce optimum conditions; check the signal operation and resulting traffic movement to ensure that conditions were optimum.

It was pointed out that a large digital computer, though costing little, if any, more than the specialized analog equipment, could perform all functions with sufficient speed to ensure that system operation would vary almost instantaneously with changing traffic demand. It was further pointed out that by simple programming the performance of any type of existing system could be duplicated, and that the mode of operation, timing, and even phasing of an individual intersection could be altered without changing equipment or even leaving -the central control site. As a bonus, a large general-purpose digital computer would be available for detailed analysis and evaluation of the results of such experiments, or for other work that might be required.

Although this suggestion appeared attractive, some doubts were expressed concerning the ability of a digital computer to perform in the required manner; therefore, authority was obtained for a small pilot project in which a network of some 15 signals would be controlled in various ways ranging from pretimed to fully traffic-responsive. This project, starting in 1959 and lasting approximately one year, proved so successful that it was decided to proceed with the full-scale system. However, many delays ensued, and it was not until October, 1962, that purchase of the necessary equipment was approved, though previously the Metro- politan Toronto Traffic Authorities had been granted, by Act of the Provincial Legislature, authority over all existing or new signals within the area. This act was necessary since a u s e d system could only be operated by a single authority and jurisdiction in matters concerning traffic control signals had previously been split between the metropolitan government and the thirteen area municipalities.

Delivery and installation of equipment commenced in Spring, 1963. The computer itself arrived in June and became operational in August, when the first small group of

PROCEEDINGS OF THE IEEE, APRIL 1968

TABLE I METROWLITAN TORONT(~STATISTICS

Item : 1957 ~ 1964

Area (square miles) 240 240 Roads and streets (miles)

Expressways I 0~ 15 Arterial ~ 285 ~ 356 Other 2,174 2,347 Total 2,459 2,718

Signalized intersections ~ 496 712 Population I 1,380,000 1,740,000 Motor vehicle registration 453,000 639,000 Persons per vehicle 3.05 2.72 Traffic accidents

Fatalities , 125 134 Personalinjury I ~ 6,935

~ 11,731 Other 13,268 16,809 Total 20,328 28,674

Property damage due to traffic accidents $6,689,000 i $1 1,411,000 Accidents at signalized intersections 4,386 5,935 Average accidents per signalized

intersection ~ 8.84 8.34 Accident rate per 100 000 population 1 1,472 ~ 1,646 Fatality rate per 100 000 population ~ 9.1 ~ 7.7 Approximate total miles driven per year ' 2,975,000,000 ~ 4,292,000,000 Approximate average annual mileage ,

per vehicle (miles) 6,567 ~ 6,717 Approximate average daily trip per ~

vehicle (miles) 26 27 ~

~~

about 20 signals were connected in. Since that time, the development of operational techniques and the connection of additional signals has proceeded slowly but steadily to the point where over 500 signals on major arterials are now under effective control. The remaining 350 existing signals, almost all of which are on minor streets, will be connected during 1968, and new installations totaling between 30 and 40 per year will be incorporated as they occur.

A special note is in order regarding the tables and figures in this paper. They are supplementary to the text in that they provide examples and additional information rather than specific clarification or illustration of the discussion. For this reason, they are not referred to explicitly but are inserted near the related written material.

EQUIPMENT The pilot project undertaken in 1959 and 1960 was

specifically intended to show that a general-purpose digital computer could optimize conditions in a given area through the effective operation of traffic control signals. The project was not intended to develop or test such equipment and methods as might be required for the efficient and economi- cal implementation of a full-scale system, nor were the years between the successful conclusion of the pilot study and commencement of work on the overall scheme used for this purpose. This lack of adequate research and development proved most unfortunate. Since at the time the whole concept was quite new and there was little or no background experience to draw on, the original equipment specifica- tions had to be written without adequate knowledge of requirements. This was particularly the case with the vehicle detectors, and much time was lost in trial-and-error testing

HEWTON: TORONTO TRAFFIC CONTROL SYSTEM 519

TABLE I1

COST FOR FOUR HUNDRED INTERSECTIONS THE METROFQLITAN TORONTO TRAFFIC CONTROL SIGNAL SYSTEM-ESTIMATED COST TO COMPLETION BASED ON ACTUAL

,

Item ~

Initial cost in dollars , Annual

Quantity Material Labor ' Total operating cost in $

Remarks

Computing equipment complete Computer site Communications rental Telemetry equipment including central

Intersection control cabinet including electronics mounting hardware, local power supply, etc.

power supply

Signal controller modification unit Signal controller monitor logics Vehicle detectors Consultants' fees Programming staff

Computer operating staff Analytic stafT

1 1,960,000 1 30,000

4,000 prs. -

7,000 490,000

1 .000 150,000 1,000 , 90,Ooo 5.000 ' 82,000 2,000 2 78,000

3

6 9

-

- -

- 1,960,000 50,000 80,000

- 8 -

500,000 900,000

150,000 1 300,000 20,000 1 10,000

100,000 182,000 1,000,000 1,278,000

500,000 500,000

20,000 30,000

176,000

60,000

50,000 20,000 10,000

100.000

20,000

30,000 50,000

-

Fixed cost maintenance contract Including power, water, etc.

Including computer input-output wiring

Not including signal controller

Average two loops per card

Senior and two junior scientific

Senior and five junior operators Senior and two junior analysts and

programmers

six field investigators

Total 3,080,000 2.320.000 5,400,000 566,000 -

of equipment as it was actually being delivered and many, sometimes rather makeshift, moditkations were necessary to achieve a satisfactory level of performance.

Communications Lines leased from the Bell Telephone Company of Can-

ada provide interconnection between the Control Centre and signalized intersections. Four pairs are used for each intersection, with the capacity of two of them expanded to 10 channels each by simple audio-tone telemetry equipment. All pairs provide a direct copper connection, and no use is made of telephone company booster or other equipment. The central site terminal frame has facilities for 5600 pairs.

For each interesection two pairs are used in a 4-wire common-return floating ground configuration to carry control signals to the signal controller. The other two, which carry monitor and detector data back to the Centre as audio tones, are also ungrounded and are coupled on and off through small isolating transformers. Line protection is provided by standard carbon block protectors at either end supplemented by small glow lamps across the coupling transformers. Standardization of signal level to compensate for differences in line lengths and characteristics is achieved by connecting variable padding resistors across the input secondary of these transformers.

Communications-Telemetry Transmitter The telemetry transmitters are small audio-frequency

tone generators, built on individual plug-in printedcircuit cards, up to 20 of which can be housed in a controller base unit. Each generator operates continuously except when keyed off by the presence of an appropriate signal at its input terminal. Each of the two pairs carries 10 tones with frequencies, whose harmonics do not conflict, ranging from 350 to 2600 Hz.

Communications-Telemetry Receiuer The telemetry receivers, also constructed on plug-in cards,

are housed in two cabinets at the central site, designed to accommodate 3500 doublecircuit cards. Each receiver controls a solid-state switch capable of changing the voltage in the controlled circuit up or down over a 15-volt range in less than 20 milliseconds. The receivers are provided with four independent well-regulated power supplies with provision for load sharing should one unit fail.

Central O@ce Equipment-Site The central office has been recently moved from the lobby

of the Old City Hall, where it had the disadvantage of having no ceiling over half the enclosure, to more conventional completely enclosed quarters in the new police headquar- ters. The major units of equipment form an L-shaped wall within the room, fitting snugly between the raised floor and dropped ceiling whxh are used for interconnection wiring and as air conditioning plenums.

Four 54011 air conditioning units supply cold air via the floor through the equipment to the ceiling and maintain room conditions at 72' F and 48-percent relative humidity.

Power for the Centre is provided by two 200-ampere, 208-volt, three-phase services controlled by magnetic switchgear located in a small basement room which also houses the voltage regulator for the computer and the 400- cycle motor generator required by the central processor. In the event of an emergency, all power to the computer room can be shut off either automatically by combustion detector alarms or manually by strategically located ham- mer-glass switches. Central Ofice Equipment-Primary Computer Complex

The primary computer used for control and data processing functions is a Univac 1107. The central processor features :

U

w I

V

A

C

1107

C

0

Y

c U

T

L

R

TACC READ

CONTROL

L I I I I

TACC U O U T W T

L REAO/WRITC N 9 CONTROL

I V D l S T R l l U t O R

1

MODlFlCATlON ¶IONAL

UNIT bo '

C O N T R O L L L R - 0 L 1 I

0 T

C

H 0 N

L MONITOR

' t T R A f I I C UNIT a l o w A L

u \ V L W I C L C acIl30n

Fig. 1. The Metropolitan Toronto Traffic Control System-functional diagram.

HEWTON: TORONTO TRAFFIC CONTROL SYSTEM 58 1

Fig. 2. The Metropolitan Toronto Traffic Control Centre (Old City Hall site).

thin-film control memory of 128 words, each 36 bits, with an access time of 300 nanoseconds ; fifteen index, 16 arithmetic, and 36 control registers ; ferrite core memory of 32 768 words with an access time of 2 microseconds; a real-time clock with I-millisecond resolution and interrupt capability; interrupt-oriented input-output facilities capable of fully buffered transmission at a maximum rate of 250 000 words per second over 16 input or output channels ; a control console which includes a lo-character-per- second buffered type printer and entry keyboard.

Peripheral equipment directly associated with the 1107 comprises :

a magnetic drum memory with a capacity for almost 5 million characters, a transfer rate of 360 000 characters per second, and an average access time of 17 milliseconds ; six magnetic tape handlers which use a recording density of 1000 nine-bit frames per inch for a maximum transfer rate of 180 000 digits per second ; a high-speed printer whch operates at 600 lines per minute; a card reader and card punch, with speeds of 600 and 300 cards per minute, respectively.

Central Ofice Equipment-Secondary Computer A prototype Univac 418 processor, which is coupled to

the 1107 through a special intercomputer synchronizer, is used as the on-line traffic input-output device. This secondary computer features :

1) ferrite core memory of 16 384 words, each 18 bits, with

2) interrupt-oriented input-output capability over 8

3) punched paper tape reader and punch, with speeds of

an access time of 4 microseconds;

single or 4 dual (36-bit) channels;

200 and 110 characters per second, respectively;

4) a control console similar to the 1107’s; 5 ) direct access to the 1107 tape handlers.

At the present time the 41 8 processor is used in two ways, either as a slave to the 1107 or independently. In the former case, it serves as little more than a buffer for the traffic data collected and signals presented to the street equipment. Used independently, it can either implement a limited form of traffic control or record detector data on magnetic tape or initiate special function operation of remote devices. In the near future, the independent and slave modes of operation will be integrated so that the 41 8 can assume control of traffic without operator intervention should the 1107 fail or be required for other work.

Central Ofice Equipment-Input Scanner The input scanner acts as a commutator to sequentially

examine the state of each telemetry receiver switch and present this in binary form to the 418. The presence of the appropriate signal which results in turning off the incoming audio tone and opening the receiver switch is recorded as a binary one; the absence of a signal which maintains the audio tone and results in a closed receiver switch is recorded as a binary zero. Thus the occurrence and duration of a signal can be deduced from the number of one bits rec- ognized. Interval lengths determined in this way have a possible error of one scan so that accuracy is a function of scan rate. As a measure of protection from transmission transient false signals, program logic requires that two successive ones preceded by at least two zeroes occur before recognizing the start of a pulse.

Signal monitor inputs are scanned in groups of five at a fixed rate of twice per second with the results stored in specifically addressable 418 registers. The address of the data determines which signal is in question, the binary content indicates the position of the signal controller cam assembly and hence identifies the interval, and the number of scans during whch the input remains constant is a measure of the interval length with a maximum error of one- half second. Five thousand inputs can be scanned in this way, so that 1000 signals can be monitored using the present five-level code.

Vehicle detector inputs are scanned individually at a rate of 64 times per second with the results being stored in 418 memory. The number of vehicles crossing a given detector can be derived from the number of strings of binary ones recorded and the length of time a vehicle is over the detector can be estimated from the number of binary ones in any string. Two thousand detector inputs can be handled.

Central Ofice Equipment-Output Distributor The input signal monitor scanner and the output dis-

tributor use the same drive circuits and are wired in such a way that when a given series of five inputs (one signal) is connected to a specfic 418 input register, three uniquely corresponding 418 locations will be connected to a specific group of three output relays. These relays are of the bi- stable mercury-wetted contact type whose state can be changed by an appropriatc binary pulse whose duration

582 PROCEEDINGS OF THE IEEE, APRIL 1968

may vary from 50 milliseconds to infinity under program control. Contact closure is produced by a binary one in the corresponding 418 register. The three relays are designated as “hold,” “actuate,” and “special function,” to correspond to those in the signal controller modification unit.

The relays in the output distributor make or break contact between a dc voltage source and the output telephone lines to produce an appropriate response at the signal controller. A variety of voltages can be selected to provide some compensation for variations in line lengths and characteristics.

In the event of power failure or absence of valid 41 8 output all relays are automatically set to their open state, thus releasing signals from computer control.

Central Ojice Equipment-Display A large photographic map of the metropolitan area has

each signalized intersection marked by a small neon lamp. These lamps are connected in series with the output distributor hold relays so that each is lit when its intersection is under computer control. This display is not only popular with the public but also very useful to the computer operator as a visual indication of the progress of initiation or termination procedures and as a reminder of signals ex- cluded from control for road construction or special events.

Two “dummy” controllers, identical to typical signal controllers except that the lamps controlled are on a small display rather than in signal heads, are also located in the Centre for display and system-testing purposes.

Field Equipment A feature of the computer system is that any type of

existing signal controller can be utilized provided that it is capable of being manually operated. For this reason, existing controllers, which were of the basic, multipledial, pretimed type, were simply stripped of their box and ex- traneous equipment, thoroughly overhauled and remounted in a new free-standing pedestal-type box which provides separate compartments for the controller, the electronics, and the telephone line terminations. Since local control is scheduled only during periods of light traffic, a single timing dial with average condition settings is adequate and the removal of time switches, transfer relays, etc., associated with multiple-dial control, has eliminated a major maintenance problem. The vehicle detectors used at semi- actuated intersections are provided with both relay and telemetry outputs so that this type of operation can continue under either local or central control.

Field Equipment-Modification Unit

The modification unit, which permits remote operation of the signal control mechanism and other devices, consists of three relays mounted in an easily replaceable plug-in enclosure. The relays themselves are of the telephone type, with pickup and dropout times of 40 and 25 milliseconds, respectively.

The three relays correspond to and are operated by their counterparts in the output distributor to perform the following functions.

Fig. 3. Intersection control cabinet.

1) Hold: When energized, this relay stops the dial timing mechanism, breaks the normal electrical connection between it and the signal switching device, and connects the latter through a set of actuate relay contacts to either a power source or ground as may be required. The hold relay is energized continuously while the signal is under computer control, but in the event of any failure, the relay automatically opens to ensure continued signal operation under local control.

2 ) Actuate: When energized, this relay completes the circuit between the signal switch and power or ground, thus causing it to operate and change the signal aspect. In normal operation, this relay is only energized for approximately 250 milliseconds.

3) Special Function: This relay may be used to select different signal phasing arrangements, or to control such auxiliary devices as lane control signals, internally illuminated blankout signs, etc. When only one special feature is involved the relay may control it directly, being energized continuously for as long as necessary. When control of more than one special function is required, the relay must be pulsed and used to step some sort of selector mechanism.

When the hold relay is energized, the normal switch and push button provided for manual operation of the signal are disabled. Each controller is fitted with a pilot light energized by the hold relay and mounted directly beside the manual switch to warn a police officer desiring to manually operate the signal that this is not feasible unless he first calls the Control Centre to have the signal dropped from computer to local control.

To facilitate installation, maintenance, and adjustment, each modification unit is fitted with a standard telephone jack connected to one of the control pairs so that service personnel using self-powered handsets can talk to their counterparts in the Centre.

HEWTON: TORONTO TRAFFIC CONTROL SYSTEM 58 3

Field Equipment-Power Supply The local power supply is required to operate all the

electronic equipment at an intersection and is designed to provide a well-regulated 15-volt dc output to a maximum of 20 telemetry transmitters and 15 vehicle detectors, with no possibility of lesser demand adversely affecting its operation. Power supply filtering must be adequate to prevent crosstalk or faulty operation due to power line or switch transients originating in the controller mechanism itself or outside. These units are built in easily replaceable plug-in form.

Field Equipment-Signal Monitor In order to ensure orderly transition from local to central

control and to permit verification of correct signal operation under computer control, the position of the signal switching mechanism is continuously monitored.

In order to simplify programming and to facilitate changes in operational mode or phasing, all controllers are fitted with 12-segment switchmg cams and each signal cycle is considered to consist of 12 distinct intervals, each of which is uniquely identified by its monitor code. Since 12 intervals are more than enough to meet the requirements of the most complicated intersection, any signal aspect may be shown for more than one interval.

To provide standardization and further program simplification, interval one is always the green indication for the major street. Pickup and dropout of central control occur at the start of this interval, and offsets are specified in absolute terms relative to its commencement. Typical interval-aspect arrangements are as follows :

Simplest Two-Phase Case ~

Two-Phase With Advanced Green

Phase Aspect Signal Interval Phase Signal

Aspect Interval

A green 1,23394 amber 5 all red 6

amber 11 all red 12

B green 7,8,9. 10

A green amber all red

B green amber all red

A advanced green

clearance

Monitoring is accomplished by connecting selected signal terminals through resistive OR logic circuits to telemetry transmitters in such a way that when a given signal circuit is energized one or more transmitters will be turned off, thus establishing a unique pattern which can be interpreted in binary form as in interval identification code. The OR logics are built four to a plug-in printedcircuit card, ten of which can be accommodated in the controller cabinet base unit.

Power and transmission troubles, short circuits, etc., are indicated by invalid codes received ; signal switching mechanism malfunctions can be diagnosed by an improper or unchanging sequence of codes ; however, a signal lamp

failure cannot be remotely shown since the monitor information is derived from the controller output terminals.

Field Equipment-Vehicle Detectors Based on experience gained during the pilot project, and

a further three years of limited study during which almost all of the then available types of vehicle detectors were tested, it was decided that the best results could be obtained through the use of inductive loop units. The following features contributed to this decision.

Installation and maintenance costs and inconvenience are minimized for the electronic units by their simple solid-state long-life design, by their location in protec- tive surroundings remote from the sensors where service interferes in no way with traffic, and by the simplicity of the calibration process. The actual sensor loops are also economically attractive to install and maintain. Their location in the pavement protects them from damage during snow removal and minor road repair, while their actual cost is low enough to render them expendable if badly damaged. Detector output can be closure of a solid-state switch coupled to a telemetry transmitter for remote operation and/or closure of a relay for operating conventional local control equipment. Further flexibility is offered for multilane use by the separation of sensors and electronics so that single-lane loops can be treated individually or together, and either true presence or reset logic can be utilized, with the change of these arrangements at any time being simple and in- expensive. Detector operation is not critically affected by vehicle size, shape, weight, or speed, by loop material, shape, or size, or by electrical or magnetic fields in its vicinity. Counting accuracy is high and reproducible, presence detection is very reliable, and the output pulse duration is directly proportional to the time spent by a vehicle in the detection zone.

The electronic circuitry for the equipment used is built on plug-in cards with provision in the mounting base for 15 units with their associated telemetry transmitters. A single unit can handle up to 5 loops for presence detection purposes but count accuracy suffers for more than 2. Loops may be located up to 500 feet from the electronics.

Field Equipment-Vehicle Detector Location Detectors used to communicate actual traffic conditions

for an area or to report the number of vehcles to be accommodated by the green phase at the nearest downstream intersection are located approximately 300 feet back from the stop bar at major intersections. This distance is selected so that operation of the detector will not be adversely affected by normal queue formation and also makes some allowance for increased queue length due to congestion caused by turning movements.

At intersections where the signal is operating in a traffic- responsive mode the signal indications will not change or accommodate special movements unless traffic is present on

584 PROCEEDINGS OF THE IEEE, APRIL 1%8

an approach facing red and detectors must be located so that any vehicle reaching the stop bar must cross a detector. This is achieved by placing the loop adjacent to the stop bar, or, in some cases, between the stop bar and the nearest driveway or other entrance.

Detectors may also be located on major arterial approaches to the area as much as a mile distant from the nearest signal location so that the system can respond to sudden massive fluctuations in traffic volume in time to prevent congestion.

Field Equipment-Vehicle Detector Loop Size Whether or not the use of multilane detection is intended,

individual loops are installed in each lane and separately wired back to the control cabinet, so that the operation may be varied at any time by simply changing connections and removing or installing an electronic module. Loops which may be connected in parallel must be made approximately the same size so that their impedance and sensitivity are balanced.

The width of individual loops measured in the direction of travel should be such that when traffic is moving no vehicle can enter the zone of detection before another has left it, and when traffic is stopped the zone of detection should not lie completely between vehicles. The former restraint would imply a loop width of about 4 feet, the latter, about 10 feet. The Toronto system has standardized on a loop width of 4 feet for all detectors except those used to indicate the presence of a stopped vehicle on an actuated approach.

The length of loop measured across the roadway must be such that a vehicle travelling in the correct direction cannot fail to be counted and a vehicle straddling the lane line cannot be counted twice. These requirements are satisfied by a loop length about 3 feet less than the lane width with the loop centered between the lane lines or placed 3 feet from a curb.

Field Equipment-Vehicle Detector Installation Before installation of a loop, its actual position is chalked

on the pavement and close observation is made to ensure that turning or lanechanging movements due to minor streets, driveways, grades, curves, or pavement conditions will not render the detector information inaccurate.

The loops, usually 2 to 4 turns of ordinary 14-gauge in- sulated wire, are installed in a cut 1 inch deep and $ inch wide made in the pavement using a diamond-bladed power saw. The loops are rectangular in shape, and those in adjacent lanes are staggered by one loop width to minimize cutting without unduly weakening the pavement.

Connection between the loops and their electronic equipment, which is usually located in the nearest intersection control cabinet but may be separately housed, is made by TG58U coaxial cable installed in a pavement cut similar to that used for the loops, located to avoid pavement weakening or interference with markings.

All pavements cuts are filled using a hot-poured rubber- ized asphalt compound, poured at a temperature of about 400" F.

Each loop and feeder cable assembly is tuned for optimum impedance match between it and the standard electronic module by permanently wiring a capacitor across the terminals so that it becomes, in effect, a part of the loop assembly; the correct value for this capacitor is selected by using a decade capacitance box and adjusting until a specsed voltage is registered at a circuit test point.

Field Equipment-Vehicle Detector Calibration When a detector is installed so as to cover only a single

traffic lane, each vehicle passing through the zone of detection produces a discrete output pulse, the duration of which should be inversely proportional to speed for a given loop width and vehicle length. Unfortunately, this relationship does not hold quite true since a vehicle must generally penetrate the loop to some extent before on-triggering occurs while off-triggering may occur either before or after it is completely cleared. These effects tend to vary with the vehicle's magnetic con6guration and speed and with the make of detector. To compensate at least in part for these and other errors, it is desirable to calibrate selected units from each supplier.

Calibration may most conveniently be carried out by installing two loops in one traffic lane with their leading edges exactly 44 feet apart. If this work is done in the normal permanent fashion with both loops wired to the nearest intersection box, then any electronic units for test can be installed and their outputs picked up in the central office. The actual calibration consists of noting the relation between the actual pulse width and the actual speed as determined by measuring the time interval between the leading edges of the pulses from the two loops. Either a standard vehicle driven at various speeds or the normal traffic flow may be used, the latter being preferred, since the results will tend to be more realistic when applied to actual conditions. In either case, an observer should be present to ensure that each measurement is free from gross errors such as might result from the passage of large trucks or two or more vehicles in close proximity.

In the general case, it has been found that the relationship between pulse length and speed is approximately

s = ( A - BPYP where

S = speed in mi/h p = pulse length in seconds

A and B= constants whose values depend to some extent on the design of the detector circuitry.

For the equipment currently in use this relationship provides a correlation of 0.98 between true and calculated speeds, when the constants have the values

A = 11.2 B = 3.4.

Field Equipment-Vehicle Detector Performance When properly installed and adjusted, vehicle detectors

coupled through the telemetry equipment to the 418 computer can be expected to provide data from whch the


following characteristics of the traffic stream can be derived. Volume: When used in only a single traffic lane, the

number of vehicles passing during a given sampling period T seconds long can be determined by counting the number of times n that the detector is turned on; then the volume V will be :

V = 3600n/T vehicles per hour.

When operated in this way an error of approximately f 2 percent can be anticipated. When used to cover two traffic lanes there is the possibility of two vehicles entering the zone of detection in any relative relationship from exactly side-by-side to completely separate. If it can be assumed that volume and speed are equal in both lanes, but that the vehicle arrival rates are statistically independent, it can be shown that the probable error will be approximately:

1 - (VL/528OS)’

where

V is the volume in vehicles per hour per lane L is their average length in feet S is their average speed in miles per hour.

In rush hour conditions at a location where lane usage is not affected by turning movements, it has been found that the average error will be approximately - 10 percent. During off-peak hours or conditions of uneven lane usage and/or speed, the error will be unpredictable and probably large.

Speed: Since vehicular characteristics vary considerably, little if any information can be derived from an individual output pulse. However, if the lengths of the pulses received during a given sampling period are summed to give the total detector on-time t , then the average speed during the interval will be:

S = (An - Bt)/t mi/h

where

n is the number of times the detector is turned on during the sampling period,

A and B are the calibration constants.

For single-lane application, if the characteristics of the calibrating vehicle correspond to the statistical average for the traffic stream, and a random distribution of characteristics is assumed during the sampling period which is long enough to include at least five vehicles, the probable error in average speed as determined will be approximately L 10 percent. A further error may be introduced if vehicles do not pass directly over the longitudinal center line of the loop since any sideways deviation will tend to produce a reduced pulse length leading to a variation of as much as + 10 percent for a center-of-vehicle to center-of-loop dis- placement of 5 feet. With multilane detection little or no speed information is available unless a high degree of in- accuracy is tolerated, since the length of any given pulse or series of pulses can depend on vehicle spacing as well as speed.

Density: The density of traffic in a given lane cannot be measured directly, but is determined from other factors, thus :

D = V/S - 3600nt/T(An - Bt)vehiclesper mile

where V is the traffic volume in vehicles per hour S is the average vehicle speed in miles per hour t is the detector on-time during a sampling in-

terval T seconds long A and B are the calibration constants.

For single-lane applications the error in density measurement will be of the same order as the factors used in making the determination. For multilane usage the errors will be increased, and results can be regarded as only very approximate.

OPERATION Insofar as system control is concerned, the basic fun,ction

of the computer is to inspect each individual signal approximately once per second and to determine if its aspect should be changed. This is done by comparing the elapsed time for the current indication with that time considered necessary to satisfy both the needs of the alternating traffic flows and the system as a whole. In some instances the required time may be predetermined, but for other than minor intervals such as advanced green or clearance this renders the operation inflexible. Therefore, to achieve maximum efficiency, the computer must calculate the optimum duration of each interval in any of a number of different ways, depending on both the physical characteristics of the location and the actual traffic conditions. Since each location may require different treatment, the operation of the computer is governed by a control plan which specifies for each signal in turn the exact factor to be considered and procedures to be used in determining the length of any interval.

Control Plan To provide for predictable variations in the requirements

both at and between individual signals, and also to facilitate the implementation of special arrangements such as flashing operation and emergency routes, many different control plans are available either in memory or on tape depending on frequency of use. The plan actually in effect at any time may be changed either manually, by console type-in, or automatically, by a selection routine. In the latter case, the change may be initiated and the new plan selected either by a coded input from a remote location, by time of day, or by actual traffic conditions as indicated by specifically designated sampling detectors.

The criteria for traffic-responsive plan selection may be density averaged over at least 10-minute intervals and weighed as necessary to ensure stable operation. For the selection of certain features, such as flashing operation, the density considered is the maximum of that recorded for opposite directions of flow; for other purposes, such as determining the requirement for progressive movement,

586

the choice will depend on a comparison between the values for opposing flows on the common street. In the latter case the direction having the larger density will be favored, but the degree of coordination, and hence in certain cases the width of the through band, will depend on- the relative values. The numerical value of either maximum or relative density used to trigger a change is determined initially by trial and error coupled with observation of actual traffic conditions on the street in question.

Parameter List For each control plan the computer is provided with a

parameter list, which specifies for each intersection such factors as :

1) the intersection identification number; 2) the control group within which it is to be included ; 3) the absolute offset or time, relative to some arbitrary

time, at which the first interval may start; 4) the length of each of the 12 intervals, together with a

reference to the signal aspect showing during each; 5) the interval, or intervals, at the end ofwhich actuation,

preemption, or special function selection may take place, together with a reference to the vehicle detector or other device which may call for this action, and de- tails of the new interval assignments and lengths;

6) a reference to any special requirements, due either to physical conditions or traffic movement at or adjacent to the intersection, which may influence its operation.

The parameter lists are prepared by determining the average signal timing required during the period of the day under consideration. Thus if the signal is to operate on an independent basis the following will be used :

Advanced green :

1OpL/(1333 - Ve).

Green : The greater of the pedestrian requirement 5 + 0.25W

or the vehicle requirement

10 (0.7 + ~,p)/(1333 - E). Clearance : The greater of the pedestrian requirement

0.1 w or the vehicle requirement

1.5 + 0.55 + 0.7D/S.

In each case

P L

Ve

ue

is the number of signal phases in use, is the number of left turns per hour, is the sum of the maximum unidirectional equivalent volumes in vehicles per hour entering the intersection on all phases, is the maximum unidirectional equivalent volume on the phase in question,

S

D

W

PROCEEDINGS OF THE IEEE, APRIL 1968

is the average approach speed in m i l e s per hour during the phase in question, is the maximum distance in feet that a vehicle must travel to completely clear the intersection during the phase in question, is the maximum distance in feet that a pedestrian must walk to reach safety during the phase in question.

Equivalent rather than actual traffic volumes are used in these calculations to provide some compensation for variations in both physical and movement conditions, since:

V, = (V + 0.6L + 0.5H)/n

where

V, is the equivalent volume, V is the actual volume as derived from previous

L is the number of left turning vehicles, H is the number of heavy trucks, buses, etc., n is the number of usable traffic lanes.

studies or from detector input,

Signal timing calculated in this way includes allowance for driver perception-reaction time and acceleration and deceleration losses. In the case of green, there is 95-percent probability that all vehicles arriving during a complete cycle will pass during the next interval. In the clearance case, there is provision for either a comfortable stop or safe passage.

When coordinated operation is required, all signals within the related group must use the same cycle length and this must be determined on the basis of both average intersection and offset requirements based on normal two-way movement. For a small group of signals the offsets can be determined in the conventional manner, but for larger groups an off-line linear or network optimization program is used. In either case, if the cycle is at least:

T = 1333p/(1333 - Ve),

where

p is the number of phases V, is the sum of the maximum unidirectional flows in

vehicles per hour on each phase at the busiest intersection within the group,

then the length of the green intervals at all other intersections should be adjusted to :

t = 0.7 + ueT/1333,

where

v, is the maximum unidirectional phase volume in vehicles per hour at the intersection in question.

As before, the actual times used must be at least equal to the minimum pedestrian requirement. But if the sum of the required phase times is less than the cycle, then the excess time may be either distributed or given to the common street to widen the pass band.


After the desired signal timing has been determined, the interval lengths for inclusion in the parameter list are arrived at by assigning the required pedestrian time to the first interval during which a green indication is shown and distributing the balance of the green time over any other intervals during which the aspect remains unchanged. Generally, no variation or preemption will be permitted until the end of this first green interval, to ensure that the nimimum time required for pedestrian safety is included. When it is absolutely certain that there will be no crossing, or where special arrangements are made for separate pedestrian actuation and signal indications, the first green interval may be reduced. Since too sudden change in aspect can cause confusion and possibly accidents, about five seconds is considered the shortest time usable.

Two intervals are normally assigned for clearance purposes. The first, corresponding to amber, is set to three or four seconds, and the second, or all red, is allocated the balance of the required time. When advanced green indications are used, this time also will be distributed over two intervals with the second, during which the flashing indication is removed, being set at two seconds to provide a warning of change and some clearance.

Trafic Data File In addition to the parameter list, the computer maintains

a traffic data file which is continuously up-dated to show in one-second increments for each intersection or vehicle detector:

1) the elapsed time for the current interval ; 2) the presence or absence of a vehicle on a secondary

detector, the occcrrence of a pedestrian push-button actuation, or the need for preemption or special function selection indicated by manual type-in or by automatic call from a railway track switch, fire hall, etc. ;

3) the number and duration in sixty-fourths of a second of the pulses coming from each primary detector.

In addition to the above information, which is obtained and recorded by the 418 computer, the 1107 is used to further process the raw data using the calibration factors to give traffic volume, speed, and density over sampling periods of any desired length. Using appropriate weighing factors, estimates of delay and congestion can also be arrived at, and a very complete picture of current traffic conditions obtained.

For certain applications a prediction of future rather than a record of past events is required. Therefore, any or all of the incoming information may be subject to further processing using any one of a number of different techniques. The most generally applicable technique appears to be the exponentially weighed moving average, which gives :

B is the measured value for the last sampling period, C is the last predicted value, W is the weighing factor.

Operation-Pickup Routine After the computer memories are correctly loaded from

appropriate program tapes, a manual type-in at the 1107 console engages the pickup routine which automatically begins the read-in of signal monitor and other information and activates the hold relay at each intersection the second time the number one interval commences. If, for any reason, the hold relay does not respond to the pickup instructions, this is repeated in two successive local cycles; then, if there is still no response, a fail message is printed out and no further attempts made except upon direct manual instruction from the operator.

Accomplished in this way, pickup tends to be random with each intersection coming in as its timing dial reaches the appropriate point, but the whole system can be under control in little more than two local cycles which is never more than about three minutes.

Operation-Signal Control Routine The control decisions governing the operation of any

signal, or group of signals if their requirements are such that they must operate in an exactly similar manner, are made in accordance with the procedure indicated by the control plan in effect using information obtained from the traffic data file and parameter list. The general procedure, repeated continuously for each signal at intervals of approximately one second, is as follows.

1) Read-in the signal monitor code, identify the current interval, and note the time for which it has remained unchanged.

2) Check to ensure that the monitor code is valid both with regard to binary content and proper interval identification as compared with a memory of previous scans or change instructions. If the code is not valid, this may be due to line transients or other temporary troubles. There- fore, no action is taken unless the error is repeated in three successive scans, when the signal will be dropped from computer control and the operator notsed by console print-out that this action has been taken for a specific reason.

3) If the code is valid and the elapsed interval time sat- isfies the specified or calculated requirement, check whether or not any preemption is allowed or necessary, and if so, set the special function selector and terminate the interval. The need for preemption will be indicated by a special entry in the data file, resulting from manual type-in, railway track circuit switch closure, or other input, which will also cause the special function circuit to select the appropriate signal display to become effective as soon as normal operation is terminated. Once introduced, preemption will remain in

A = B + W(C - B ) effect until the instruction is cancelled either by type-in or by

where removal of the special input. In certain cases, such as railway crossings, preemption once effected may also be locked

A is the predicted average value of the factor for the in at the local controller to provide protection against any next sampling period, form of system failure.

588 PROCEEDINGS OF THE IEEE, APRIL 1968

l t

I I

d

TCRMIMATL

\

Fig. 4. Logic of signal control routine-any intersection, any interval.


4) If no preemption is required, determine whether the interval may be extended either indefinitely or to some limited extent in the absence of any demand for a change. If there is no demand and the maximum hold time, if any, has expired, then the interval may be changed immediately, but if the signal is not actuated, or there is demand, the actual interval length required must be computed in accordance with any one of several special routines.

5) Determine the required interval length; then issue the change order when the computed and elapsed time equate, having first checked the need for any special phasing arrangements to be introduced at the start of the next interval, and, if necessary, having set the special function selector. Unless the control plan contains contrary instructions, a request for special function selection will be automatically cancelled as soon as the required arrangement comes into effect.

Operation-Interval Length Routine For a predetermined mode of operation the exact length

of all intervals is specified in the parameter list. For traffic- responsive operation, where interval duration may vary, only minimum values are given and the need for extension beyond this is determined by a special subroutine. A great many different time-determination procedures may be used either for different signals, for the same signal at different times of the day, or for different traffic conditions. For any signal, the actual procedure to be used at any instant will be specified by the control plan in effect.

In the general case, variations are only required in the length of the normal green time, but in certain instances, the advanced green time may also be made responsive to traffic demand. In the simplest case, this variation may be accomplished by assuming that the turning movement is a fixed percentage of the through traffic and varying the allocated time with the approach volume. More effective operation will result where separate lanes and turning movement detectors exist, since the variation can then range from zero in the absence of demand up to whatever maximum may be set, approximately two seconds being allocated to each vehicle.

In some special circumstances, usually when the intersection is very wide, it may be advantageous to make the length of the clearance intervals dependent on actual approach speed or to shorten them considerably if there are no vehicles between the primary detector and stop line. In the former case, the required time may be determined by insert- ing actual speeds in the formula previously given, while in the latter case, the amount of reduction will depend on local conditions.

Where all signals operate independently of each other the length of the normal green indications can be calculated in any way desired, including predetermined, semi- and fully actuated, and fully traffic-responsive using vehicle density, delay, etc., as the criteria instead of, or in addition to, volume. Any of these factors may be used in the form of either recorded values from the past cycle, average or exponentially smoothed values over several past cycles, or

predictions of the probable values for the cycle in question. Any conventional or experimental relationship between the factors and time can be used, but in all cases the actual operation must be carefully observed under different traffic conditions and weighing factors adjusted as necessary to accommodate variations in local requirements.

Where the operation of signals on a given street or in a given area must be coordinated to permit progressive movement, the determination of interval lengths must be made in accordance with modified routines designed to accept the limits imposed by the required cycle and offset relationship. These latter factors may be determined either in advance or, preferably, by a special subroutine which examines the overall traffic situation and reaches the necessary decisions. Any number of procedures may be required, such as calculation of the conventional speeddistance relationship together with relative lane density, recognition of platoon movement with an estimate of arrival time, or optimization of delay along a route or around a network.

To prevent instability, the signal coordination subroutine only operates at intervals of 10 or more signal cycles to establish the requirement for the next similar period using either the record of past events or a prediction of future events. When the necessary cycle length is determined, the sum of all intervals is made to conform by imposing an upper limit on the sum of green which already has a lower limit in the sum of the necessary minima. Within these limits, the control routine in use can distribute the time between phases in any desired way using the same factors as for independent operation. Recognition of approaching platoons or estimation of relative delays may trigger a return to common street green to provide variation in offset, or this may be predetermined with any necessary change introduced over a number of cycles by lengthening or short- ening green intervals as required to obtain the correct relationship in minimum time.

Operation-Drop-Out Routine On receipt of either a manual type-in or an operational

plan instruction to terminate system control, the drop-out routine is actuated and immediately commences to adjust the operation of each group of related signals towards a predetermined average offset relationship with a common cycle equivalent to that provided by the local timing dial. When the correct relationship has been established, the hold relay of each signal in turn is released immediately following the commencement of each number one interval, thus releasing the timing dial and permitting operation to continue without interruption. Following drop-out, the operation of each signal is monitored for approximately 5 minutes and any deviation from normal indicated by a console print-out.

Accomplished in this way, drop-out may take up to 15 minutes to complete but barring local power failure or mechanical difficulties the signals will hold in the correct relationship for several hours, thus ensuring at least some continuance of service during periods of light traffic. Because of the time requirement for normal drop-out, the

590 PROCEEDINGS OF THE IEEE. APRIL 1968

procedure is of no value in an emergency; thus, in the event of central power or computer failure, all signals will almost instantaneously revert to local control and the relative offsets will be completely random. To prevent this, in certain special cases, a simple form of single-wire local interconnect has been retained between some groups of closely spaced signals.

Operatwn-Analysis Routines A feature of the system is that all relevant data concerning

both its operation and actual traffic conditions is available at the computer where it can be displayed for immediate study or stored on magnetic tape for later detailed analysis. This data tape provides a record of signal operation referenced to time which can be invaluable in certain circumstances, since under Ontario law a municipality can be held responsi- ble for damages arising from an accident which can be proved to have occurred because of a signal defect which was not corrected within what the court considers a reasonable time.

On-Line Analysis: Aspect changes at any one signal or vehicular movement past any one detector can be displayed as they occur on the 418 console type printer, thus providing useful information for service personnel engaged in check- ing equipment. This information is available even when the system is not under computer control, but when the 1107 is on line then more detailed real-time data may be obtained via the high-speed printer using either of two special routines called, respectively, RECORD and SENSOR.

The RECORD routine analyzes and prints out information from any 4 signals together with any 16 detectors to show, at one-second intervals, actual clock time, signal aspects, the number of vehicles passed during the previous second, and the length of the pulse produced by the last vehicle.

The SENSOR routine is designed primarily for testing the acceptability of detector information. It runs at all times when control is in effect, and screens the input to the data table by performing a number of tests and automatically re- jecting anythmg of doubtful validity. The tests include checks for complete or intermittent failure and for erratic performance based on unreasonable volume changes or count frequencies together with a comparison between his- torical information and the nearest up- and downstream detectors. To provide an immediate record of traffic conditions and, more importantly, to indicate the need for servicing, the results of these tests for all detectors can be printed out at 15-minute intervals together with a cumulative list of failures and both average and exponentially smoothed sampling period volumes. This print-out is generally obtained during the first hour of control each day and immediately examined so that defective detectors can be serviced with minimum delay.

Of-Line Analysis: Various routines are available for off- line analysis, or in some cases paralleled analysis, of the information stored on the data tape during previous control periods. The most useful of these routines are:

1) coums-which gives for specified detectors and time periods the average and cumulative volumes together with

exponentially smoothed volume and average pulse length if these are required.

2) PATTERN-which gives for specified detectors and time periods the average volume, pulse length, and density. The specified time periods may be successive on one day or they may be the same on different days in one week or on the same day in several weeks.

3) GRAPH-which gives a simultaneous graphical rep- resentation using different symbols of any 4 of 11 different t r a c characteristics obtained from the same or different detectors.

4) SmmM-which gives a space-time chart showing the actual relationship between the start of common street green at a number of specified signals.

5 ) corn=-which summarizes traffic data on a cycle by. cycle basis at any specibed intersection and applies various tests to indicate congestion. A modification of the routine may be used to obtain simultaneous indication of congestion at up to two hundred detector locations.

6 ) DELAY-which indicates for a specified intersection approach over given time periods the probable length of queue at the start of green and the average delay.

7) LINKswhich indicates for a specified intersection approach over given time periods the mean arrival rate, time, and delay referenced to the signal aspect in 2-second intervals.

8) TABLFS-which prints out for specified intersections the contents of the parameter list referenced to all control plans.

9) P m s w h i c h determines the presence of congestion on any or all intersection approaches by analyzing the volume-speed relationship. The result may appear as the percentage of congested cycles during a given period, thus providing a quick check on efficiency, especially when changes in operation are introduced.

To provide for long-term data storage and more convenient preparation of weekly, monthly, or yearly com- parisons the detector information contained on the original data tape is averaged over specified intervals, usually 15 minutes, and entered on a new tape which may cover a period of 200 days or more. This compressed data tape may be analyzed at any time and includes all information except that concerning signal aspect. The original data tape is normally cleared and reused after one year unless some court action is pending.

PERFORMANCE In some respects, the very flexibility of the computer sys-

tem presents a problem because there may be a tendency to introduce special effects more for novelty than utility or to forge ahead with the more sophisticated control modes without first trying the simpler methods or preparing a proper basis for comparison and evaluation. For this reason, and also to provide adequate time for staff training and equipment testing, it was decided that on their initial connection into the system all signals would be operated on a pretimed basis which would later be optimized and later still converted to traffic responsive.

+

Fig. 6 . Example of traffic analysis routine GRAPH. \o

%


Development of the system in this way has produced a considerable amount of information, but it has also slowed down the project since the available staff was not large enough to proceed simultaneously with the pickup of newly connected signals and with upgrading. This was especially the case insofar as route control was concerned. Though much information, in the form of space-time diagrams and flow charts, was available from the computer, detailed evaluation required that at least ten floating-car- type speed and delay runs be made in each direction on each street both before and after any change in signal operation. With initial pickup nearly complete and early equipment problems, especially with vehicle detectors, overcome, it is expected that development will proceed more rapidly in the future.

The computer is currently controlling signals for a minimum of 15 hours each day, 7 days a week, with the remaining time being used for maintenance, traffic analysis, and general data processing work on behalf of other depart- ments. It had originally been expected that this general data processing could be carried out on a time-sharing basis, thus considerably reducing the system costs. This has proved impractical and the work, which occupied some 2600 hours of exclusive machine time during 1965 with a peak demand of some 450 hours in one month, will be terminated as soon as present commitments are completed. When this occurs system operation will be extended to cover at least 20 hours per day.

Performance-Area Control

On almost every street there is at some point either a wide separation between adjacent signals or a major intersection whch creates a natural discontinuity for through traffic. On the other hand, there are many areas in which signals are so closely spaced or traffic conditions are so similar that close coordination and a more or less identical mode of operation is mandatory. Combining these two factors has led to the delineation of some 60 so-called control areas, each of which can be considered as a more or less independent unit within which signals can be operated as required without reference to conditions in adjacent areas except in special instances. Many of these control areas include signals in network formation such as the downtown grid whle many others include only those on a single street, or, in some cases, on a section of a major arterial route.

The control-area concept has simplified programming, data handling, and evaluation, while increasing operational flexibility, especially in the case where adjacent areas include signals on a major route, since the points of discontinuity can be used to introduce different cycle and/or offset rela- tionships. Thus special events or the normal variation in time between the commencement of rush hour at widely separated locations can be easily accommodated.

Insofar as is practicable, both the initial connection of signals into the system and all future development work is carried out on a control-area basis. This permits the available staff to be used to best advantage and also creates good

public relations because the areas are relatively small and any improvement in conditions readily noticeable.

Performance-Route Control To provide a thorough check on equipment and a basis

for later comparison, the initial operation of each group of signals on a common street was in a strict pretimed progressive mode using plans prepared by the computer, guided by an off-line linear or network optimization program and existing traffic data. Not surprisingly, a comparison between this mode of operation and the previous non- coordinated arrangement showed a very distinct improvement. Travel time and the number of involuntary stops decreased over a large area by an average of some l l and 45 percent, respectively, while the average speed and number of vehicles passing in a given time increased by some 13 and 10 percent, respectively.

Several plans designed for a.m. and p.m. peak and for normal day conditions were available; as expected, it was found that, while the pretimed operation was extremely efficient as long as drivers performed in the way they were supposed to, it could result in gross inefficiencies if for any reason conditions differed from normal. This is particularly the case during periods of extremely heavy traffic. If vehicles either left over from a previous platoon or having entered the street between signalized intersections are waiting at the stopbar, then the first vehicle in any approaching platoon may be delayed and consequently the last cut off. This effect easily becomes cumulative, with more and more vehcles from successive platoons being cut off, and thus the queue grows and serious congestion develops quite rapidly.

During the next stage of development the same predetermined plans were used, but instead of being placed in operation strictly according to time of day, they were automatically selected for use on a traffic volume basis. Though more flexible than the previous arrangement, this was still not satisfactory, especially during periods of extremely heavy or light traffic. In the former case the problem was the same as before though it was noted that bringing in the peak hour plan at the time required by conditions tended to delay the onset of congestion and consequently to shorten the peak problem periods, illustrating that it is much easier to prevent trouble from developing than to get rid of it once it has developed. In the second case the erratic behavior of drivers and the lack of platooning tended to reduce through band effectiveness even on those few streets where signal spacing was optimum for two-way progression.

In the later stages of development the original predetermined plans were still used as a basis, but they were made considerably more traffic responsive in that within program limits, the cycle and offset could vary in a predetermined manner with overall demand and predominant direction of flow, and the splits could vary with individual intersection requirements. A comparison between this and random operation showed that average travel time and number of involuntary stops had decreased by some 8 and 37 percent, respectively. The apparent decrease in efficiency

594 PROCEEDINGS OF THE IEEE, APRIL 1%8

4~ 2o 1 e a U > U

W

e 7 5 n w

L

T R A V E L T I Y E A

I -

LEGEND

SIGNAL OPERATION: 0 RANOOY CO-ORLMNATEO PRE-TIMED CC-ORDINATE0 TRAFFIC RESPONSIVE

CWROINATEO TRAFFIC RESPONSIVE AOJUSTEO TO IMPROVE CROSS STREET FLOW

ACCIDENTS ON SIYILAR STREETS: WITHOUT SYSTEY CONTROL =WITH SYSTEY CONTROL

NOTE THE ABOVE FIGURES ARE DERIVED FROY STUDIES CARRIED OUT ON SONE 4 0 Y I L E S OF ARTERIAL STREETS ! HAVING AN AVERABE OF 4 SIGNALS PER MILE.0URING THE PERIOD COVERED BY THE STUDIES TRAFFIC I

VOLUMES THROUGHOUT THE AREA INCREASED BY SOYE 9.6 X . i

Fig. 7. Overall effect of different modes of signal operation on traffic conditions as shown by floatingcar-type speed and delay studies and accident analysis.

TEST PLANS

I OFF PEAK

T

i . .

! I -mwoclrrru

- - - - TEST

I.

..

I

~ . -

a.

!

I #

i i j I

!

- I M O I5Jo 1 6 0 0 Ibu) ITW m le00 W o loa0 - 1 1 I I I I I I I I I I I I I I I I l l I I 1 1 1 1 I l l 1 1 1 1 1 1 1 I I I I I

SuANcEb H€AVY. O U r W N D #&%&A= Wrswno MONRATE, OUIBWNO M L A H C l F D CFF PEAK P M RUSH HOUR PLAN PDST-PM F#K P W 4 PRE- P.M. PEA)( PLAN OFF PEA&

I l l 1

NORMAL PLANS

Fig. 8. The effect of a plan change malfunction on peak hour intensity at a point on an urban arterial street.


between this mode of operation and the coordinated pretimed mode is explained by the fact that cross street movement has been improved and serves to emphasize the point that it is a relatively simple matter to achieve good results on the initial implementation of system control, but thereafter the law of diminishing returns applies.

This type of traffic-responsive operation proved reasonably satisfactory in that the duration of peak hour congestion was considerably reduced, though not eliminated, while the increased flexibility allowed the system to automatically adjust for variations in traffic demand. In spite of these gains, it is felt that the present mode of operation requires too much preplanning. Future work will be directed towards increasing the self-determining capabilities of the computer, especially with respect to the peak hour congestion problem, where it is possible that a solution may lie in switching from a pure progressive to a simulated volume density mode at some critical stage in the build-up period. Attention will also be concentrated on refining the progressive modes of operation by making the offset determination automatic in accordance with either actual or critical speed. The former will probably be used during periods of light traffic, since it will allow more adjustment for varying conditions, while the latter will be used in peak periods in an effort to achieve maximum capacity.

In carrying out the investigation of route control methods it became obvious that, insofar as periods of light traffic were concerned, one of the major problems was that the close spacing of many minor signalized intersections pre- vented two-way progressive movement at any reasonable speed. To at least partly overcome this problem, secondary detectors were installed and a number of these intersections were operated in a semi-traffic-responsive mode. The yield point was determined by the through street requirements, and the minor street time was either fixed at a pedestrian minimum value or allowed to increase from this in accordance with demand. This arrangement proved reasonably satisfactory during daytime, provided that the minor flow was sufficiently small that actuation did not occur every cycle. Under late night and early morning conditions, however, difficulty was still experienced in that the signal operation still caused undue delay to vehicles on both streets. This problem in turn was overcome by arranging for at least every other minor signal to be operated in a flashing mode with red showing to the side street and amber to the major. This type of operation has proved most effective in reducing delay and does not appear to have produced any increase in accidents.

Though flashing operation is presently introduced by the 418 computer on a clock basis, it is hoped to replace this shortly with a method of automatically determining the need, probably using side street volume and availability of major street gaps as criteria. Both semi-traffic-responsive and flashing operation can be used whether or not the system is under computer control, since in the former case the detectors, and in the latter case, a local time switch, can be connected through the hold relay back contacts to the local controller mechanism.

Operation-Critical Intersection Control When studying conditions at major intersections on

routes along which the signals were operating on a pretimed progressive mode, it was noted that a supposedly steady traffic flow was in reality very uneven if considered in successive short time intervals. This probably occurs because of differences in driver reaction time and vehicle performance and indicates that signal timing based on average volume considerations may result in inefficiency. To overcome this, it was arranged to retain a fixed cycle length but allow the proportion of green time allotted to either phase to vary almost directly with instantaneous demand. It was found that with this mode of operation all the advantages of progressive movement could be retained, and during peak periods the average delay per vehcle could be reduced by some 27 percent without adversely affecting platoon movement.

Some major arterial intersections are sufficiently isolated that they cannot be considered as part of a progressive system on either street, and in these cases best results were obtained using a simulated volume density mode of control with both cycle length and split varying in accordance with the almost instantaneous demand. With this type of operation it was possible to reduce the average delay per vehicle to about 30 seconds, or some 10 percent below that occa- sioned by optimum pretimed operation, while handling a peak volume equivalent to some 1400 vehicles per lane per hour of green.

With both these modes of control it was found that serious trouble could develop if volume alone was used as the basis for split variation. If for any reason congestion developed on one approach and not on the others, then in a given sampling period the detectors might indicate that the street on which free flow was taking place was carrying the larger volume and hence it would be given the larger share of green time. This process can be cumulative until the congested street is actually receiving its minimum allowable green in spite of its urgent need for more time. To overcome this, a congestion identification routine whch is triggered by a specific pattern of detections considered in relation to signal aspect was tested. This routine provides compensation by artifically increasing the count on the affected street, and in addition to this, it is intended to make split variation depend on density rather than volume and also to provide a program means whereby other detectors on each street can be periodically checked to ensure that conditions are normal.

Performance-Turning Movement Control At a great many intersections where turning movements

present a problem but are not sufficient to warrant a completely separate phase, conditions have been greatly improved by using a split phase arrangement in which the green for one direction comes on in advance of that for the other. During this usually short interval, the green for the favored direction is caused to flash rapidly thus alerting drivers both to its presence and duration while at the same time allowing the feature to be omitted at any time without

596 PROCEEDINGS OF THE IEEE. APRIL 1968

F

60

5 0

4c

QUEUE VEHICLES

PER L A N E

3c

20

10

0

2600

2 5 0 0

2 4 0 0

2 3 0 0

2 2 0 0

2 1 0 0

U 2 2 0 0 0 I IL

? 1900 In W : 1800 X Y

> 1700

I600

I500

A L L APPR TOTAL

v o

LEGEND

a PRE-DETERYINED SIGNALOPERATION

0 SlONAL OPERATION CO-ORDINATE0 TRAFFIC RESPONSIVE

U Y E C O N B E S T E D C Y C L E S

:ig. 9. Overall comparison between the effect of predetermined and coordinated traffic-responsive signal operation.

LEGEND - PRE-DETERMINED SIGNAL OPE RATION _-_- CD-ORMNATED TRAFFIC RESPONSIVE SIGNAL OPERATION

n 0 D 0 - n .a c 0

0

6

0 a 0

Q 0

Fig. 10. Comparison between the effect of predetermined and fully trafkresponsive signal operation as shown by the measurement of average queue length at a single intersection.

40

2 0

DENSITY

VEHICLES

PER

N I L E

10

0

LEBEND

- PRE-DETERMINED SIGNAL OPERATION _ _ _ _ CO-OROINATED TRAFFIC RESPONSIVE SIGNAL OPERATION

Fig. 1 1 . Comparison between the effect of Predetermined and fully tr&c-responsive signal operation as shown by the measurement of average approach density at a single intersection.


the need for special signs other than those of a simple ad- visory nature such as “advanced green when flashing.” To provide for clearance and increased safety the flashing green is changed to a steady indication for about two seconds before the opposing direction is allowed to move.

The length of the advanced green indication is directly under program control while the favored direction may be changed through the special function circuit, though presently only on a predetermined time basis. It is hoped that this selection can shortly be made in accordance with traffic demand at least at those intersections where separate turning lanes are provided and detectors can be located to record the movement. When this is done, an advance green interval of suitable length to accommodate the waiting vehicles will be provided for whichever direction requires it most, provided that the turning vehicles exceed about 10 percent of the opposing through volume. Where this requirement is satisfied on two opposed approaches simultaneously, then the feature may be omitted completely or accorded to whichever direction observation has indicated as experiencing the most difficulty.

It is sometimes desirable to prohibit turning movements where the number of vehicles involved is small or where adequate alternate routes are available. For this purpose, the special function circuit is used to control blankout- type, internally illuminated signs displaying the appropriate message. Though currently operating on a predetermined time basis, this arrangement can easily be made traffic responsive if by-laws can be changed to permit this, since at present they require that the times during which the restric- tion is to be in effect be specified.

Performance-Pedestrian Control Unfortunately, there is as yet no positive way of determin-

ing the presence of a pedestrian who may desire to cross on the green signal; therefore, a prime operating rule is that the minimum time allocated to any phase must be that required for safe passage at normal walking speed. At relatively busy intersections t h s presents little problem since the vehicle demand will generally be sufficient to utilize the time, but in other cases it may lead to serious inefficiency. To avoid this, it is necessary to install “WalkDon’t Walk” pedestrian signals and to arrange through the special function circuits that the former will only be illuminated following a push-button actuation, which will also cause both it and the vehicle green to be shown for the appropriate time with the increase over normal being either taken from the major movement or added to the cycle depending on circumstances.

At major intersections where continual conflict exists between turning vehicles and pedestrians, special arrangements have been made to facilitate the latter either through an exclusive “Walk” phase or through a split phase type of operation in which vehicular movement is controlled by directional arrow signals. In the latter case through movement can be permitted to continue while turns are tem- porarily prohibited. In both cases pedestrian “Walk/Don’t Walk” signals are used and the special phasing arrange-

ments may be either a permanent feature of the operation or they may be called in only on push-button actuation depending on the probable demand.

At minor intersections where flashing operation is used at some period of the day, there may be a problem if major street speeds are high and the pedestrians who may wish to cross are elderly or disabled. In a few instances, this has been overcome by keeping the “Don’t Walk” signal illuminated constantly and arranging that on receipt of a push-button actuation flashing operation will be suspended and major street green shown for a period of about fifteen seconds following which the signal will cycle in the normal manner. If no further push-button actuation is received flashing operation will resume after about two normal cycles.

Any of these arrangements intended to facilitate pedestrian movement are operational whether or not the signal is under computer control since the push buttons can be connected directly through the hold relay back contacts to the special function circuit and thus can actuate it. Under local timing flexibility is lost but safety requirements are satisfied.

At certain specifically marked but not signalized locations, pedestrians are granted the right-of-way over vehicular traffic provided that they wait for a gap before starting their crossing. Naturally, this latter requirement has been lost sight of and pedestrians take it that they can stop traffic whenever they so please which has created problems on arterial routes since the planned progressive movement can very quickly become disorganized if many such uncon- trolled stoppages occur. To overcome this many of these special crossings are being replaced with signals located either midblock or at a convenient intersection while many others are being abolished since with more efficient traffic control, platoon movement is encouraged and the number of reasonably safe crossing gaps increased.

Performance-Lane Usage Control On one arterial street where for a distance of approxi-

mately two miles there are five traffic lanes and a very distinct tidal flow, arrangements have been made to vary the direction of movement permitted in the center lane. This is accomplished by means of special signals installed directly over the lane at intervals of approximately one thousand feet to indicate by means of a green arrow or a red x the permitted or prohibited directions of movement. These signals are connected together in such a way that it is im- possible for confhcting indications to be shown simultaneously. They are controlled by selector relays mounted in a convenient intersection signal control cabinet and operated by the computer through the special function circuit. The selection relays are provided with a built-in time delay to ensure that on lane reversal there will always be approximately a two-minute clearance interval during which movement in both directions will be prohibited. To date, these lane usage signals have been operated on a predetermined time basis, but it is expected that they will soon become automatic in accordance with traffic demand. To accom- plish this the bidirectional traffic volume will be compared

598

and the larger favored provided that the smaller can be accommodated in the available lane or lanes making due allowance for parking, etc., which is not always prohibited.

Performance-Accident Control A comparison of police statistics for two reasonably

identical downtown sections of the city, each approximately two square miles in area with some ninety signalized intersections, has shown that in one where no change was made in signal operation, traffic accidents increased by about five and one-half percent over a two-year period, while in the other, where system control was introduced at the end of the first year, there was a decrease of about seven and one-half percent. A slightly more detailed comparison of conditions on three major arteries showed that over a length of approximately two miles, accidents have been reduced by some sixteen percent though the volume of traffic has increased by some twelve percent. Perhaps predictably, the least decrease of some three and one-half percent has been at signalized intersections, while the greatest of some six and one-half percent has been at midblock locations where rear-end collisions are most common.

Though these figures are regarded as tentative, and subject to verification in future years, they do substantiate the hypothesis that many traffic accidents result from a driver’s subconscious or even conscious reaction to delay. If this is indeed the case, then perhaps the emphasis in future safety campaigns should be on gaining freedom of movement rather than on tightening controls and restricting speed.

Performance-Special Features In a number of cases, circumstances required some special

mode of signal operation to achieve either safety or efficient results, and in handling what would otherwise have required complex and expensive local equipment, the computer system has proven its versatility. Among such special arrangements the most common currently in use are:

Snow Plum : It has been found that during even light snow signal efficiency drops sharply simply because of the time required for movement to start on slippery pavement. On the theory that once movement has started it should be allowed to continue for as long as is reasonable, and that there is not too much point in bothering about progression, special plans have been designed to provide considerably longer than normal signal intervals especially for any approaches on grades, etc. Though presently selected for use manually on the basis of observation, these plans have proved extremely useful, and it is hoped eventually to utilize some forms of sensor to automatically introduce them when conditions so warrant.

Variable Phasing, Preemption, etc.: By providing more than one signal switch camshaft and utilizing special function circuits to determine the one actually controlling the signals at any given time, it is possible to select which of a number of different phasing arrangements shall be used during any cycle. The arrangement has been used with advantage in the following cases :

PROCEEDINGS OF THE EEE, APRIL 1%8

Where an intermittent minor movement which cannot safely take place in conjunction with others must be accommodated at a complex intersection. Where a signal immediately adjacent to a fire hall may seriously impede the movement of emergency vehicles. In this case, the special arrangement is called for by a switch activated from the fire station and usually consists of either all red indications or the fairly rapid selection of the most favorable green. The special arrangement remains in effect for a predetermined time suflicient to permit any emergency vehicles to clear. Where a signal immediately adjacent to a railway grade crossing may be superlluous or even dangerous when a train is passing. In this case the special arrangement is called for by track circuit switches installed by the railway company and usually consists of an almost immediate display of red to the af€ected movement accompanied in some instances by green arrows indicating that turns only may be made. Normal green, or sometimes green arrows with those indicating turns into the affected approach omitted, may be used on the other street. The arrangement remains in effect until the train has passed.

Performance-Economic Benejits It is known that in the past year property damage result-

ing from traffic accidents amounted to approximately $1 1.5 million. Therefore, if the present trends continue, it can be said with reasonable justification that system operation will save some $1.5 million annually. These figures do not take into account such items as doctor’s or lawyer’s fees, hospital- ization, court awards, and lost time, which there is no simple way of assessing, but which must amount to a very considerable sum.

It is not known exactly what economic loss can be at- tributed to delay, but if it is assumed that half of the registered vehicles make two half-hour trips each working day, and that their operating cost, including driver’s time, fuel, insurance, depreciation, etc., is about $2 per hour, then the reduction in travel time resulting from system operation probably will save some $18.5 million annually. This figure makes no allowance for the increased operating costs of commercial vehicles or the time lost by passengers on transit vehicles, which, if they could be assessed, would un- doubtedly increase it greatly.

On a very conservative estimate based on the results achieved using the least efficient modes of signal control, the system will probably save the community some $20 million annually, and this for a capital investment of some $5 million. To have achieved the same result from the construction of new or the improvement of existing facilities would have taken a great many years and cost, in interest charges alone, very much more.

An indirect economic benefit has been that police officers who were previously assigned to manually operate signals at busy intersections have been relieved of this onerous duty,


thus permitting more work on street patrol, etc., without requiring additional men. In addition, system operation has at least postponed the need for major road construction works, thus freeing capital for other purposes.

Performance-Operational Eflciency Three years’ experience has shown that the reliability

of the computercontrolled system is in many respects better than that of the original equipment. With 3 to 4 hours maintenance each week, down time due to hardware trouble amounts to less than one percent, of whch a large proportion has been due to external factors such as power or water supply failures, the latter of which causes the air conditioning to shut down.

After some initial difficulties the vehicle detectors and other electronic equipment have also shown good reliability with only a minimum maintenance requirement though difficulty has been experienced in keeping the sensor loops intact due to the large amount of road and other construction taking place. Fortunately, loop failure is easily detected by the sensor analysis routine and repair or replacement can be effected reasonably quickly and at low cost by simply splicing in as much new wire as is necessary.

Reliability of the telephone circuits has proved extremely good though some difficulty has been experienced following heavy rain. Originally a certain amount of damage was caused to the electronic equipment by high-voltage transients due to lightning, strikes, etc., but this has been overcome by the provision of fast-acting protectors at each end of the line.

Simplification of the local control equipment and especially the removal of time switches and transfer relays has greatly‘ reduced the incidence of signal failure, while the duration of any such failure has been noticeably shor- tened since almost any defect apart from lamp burnout can be noticed by the computer through variations in either monitor or detector inputs. In many cases repairs have actually been effected before notification of trouble was received through the normal channels. In the same way, it has occasionally proved possible to notify the police of an accident, before they received a call from the scene.

FUTURE POSSIBIL~~IES The operational possibilities of a digital computer con-

trolled traffic signal system are limited only by the imagina- tion of its designers and the availability of money necessary for their accomplishment. Some of the future applications are as follows.

Expressway Surveillance and Control Expressway surveillance has so far been attempted only

on a very limited scale. A few detectors were placed at the metropolitan boundaries to indicate any variation in the incoming traffic volume, to permit the appropriate sections of the signal system to be adjusted well in advance of any sudden demand. This operation was sufficiently successful that, in the future, additional detectors will be located at

suitable intervals along all expressways and on their exit ramps to indicate, not only the amount of traffic to be expected on the local streets under normal conditions, but also, and more importantly, the onset of any trouble on the expressway itself which may necessitate special action if delay is to be kept to a minimum.

Detection of movement difficulty on the expressway will probably be based on the repetitious determination of volume-density figures at each detector station, since variations in these can indicate trouble provided that their actual values are at or above the critical value. The potential seriousness of the trouble can be judged by comparing the volume passing the station in question with that passing one or more upstream stations. The amount by which the latter exceeds the former will determine the congestion and delay likely to result and what action, if any, should be taken.

When the anticipated trouble is not expected to prove serious, blankout-type variable message signs located ahead of each detector station may be used to advise drivers to use caution, to reduce speed to a specified figure, to change lanes, etc. These signs may also be used to advise of special driving conditions which may lie ahead.

When the anticipated blockage is expected to be serious, similar signs located ahead of exit or entrance ramps may be used to warn drivers either to leave or not to enter the expressway and also to suggest the best alternative route. As these signs are activated the signal system would be adjusted to handle the expected demand.

Detour Control Blankout-type variable-message signs could be installed

at major intersections or other appropriate points to warn drivers of conditions ahead, and the need to detour in the event of severe congestion, accidents, f h s , construction, etc. With the computer’s overall picture of the traffic situation, drivers could also be advised of other factors such as the most desirable alternative route, the optimum speed for fewest stops, etc. To be properly effective, the signs would need to be completely flexible as to message content and might, therefore, be of the photoconductor-controlled electroluminescent or similar type capable of accepting a digital input and converting this into a self-storing, clearly visible legend.

Emergency Route Control If special routes were established for emergency vehicles

proceeding from various fire halls, police stations, or hospi- tals, to different sections of the metropolitan area, it would be a comparatively simple matter to provide an assured green path through the use of special control plans. These plans could be retained in memory and almost instantaneously placed in effect either manually by the operator on receipt of a telephone call, or preferably automatically by a coded message transmitted from the police emergency switchboard as the vehicles are dispatched. An even simpler matter would be the provision of special plans designed to establish evacuation routes in the event of an attack alert.

the metropoliton toronto traffic control signal system

Documents