New pedestrian facilities

DRIVE PROJECT VI061: PUSSYCATS

R-92-55

Dr. P.B.M. Levelt Leidschendam, 1992

-SUMMARY

This report is the Dutch part of an international (French-British-Dutch) evaluation study of new pedestrian crossing facilities, summarized under the name 'PUSSYCATS.

Dutch pedestrian signals consist of a red light (standing man) above a green one (walking man) positioned across the street. Before the green light changes to red, it flashes for a short period. Pedestrians may still start to cross while it does so. Red means: "If you are on the crossing, move to the kerb as quickly as possible" and otherwise "Do not cross".

Some signals change automatically, while at others, the green light can be called up using a push button control. In this case, the delay mayor may not be regulated by the presence of traffic at the crossing. Some cross-ings at junctions are 'conflict free/, while at other crosscross-ings, vehicles turning right and left may continue when the pedestrian light is green, giving way to pedestrians.

In the Netherlands new traffic regulations took effect in 1991. These regulations (RVV) include the introduction of new pedestrian signals, which traffic departments can use to replace the old type. The new alter-native pedestrian signals consist of a flashing yellow light above a green one. The flashing yellow light means: "You may cross at your own risk" . The crossing must be 'conflict free' when the light is green·

PUSSYCATS is a new system, characterized by technical improvements better adapted to the behaviour and needs of pedestrians, particularly those of vulnerable road users. The pedestrian display has been moved to the near side of the crossing (the Maastricht position), facing the oncoming traf -fic. A mat detector replaces the push button, with infrared sensors detec' ting the presence of pedestrians on the crossing.

These technical improvements make it possible to show the pedestrian green light for short periods, to cancel unused calls, and to adjust the clearance time for slow pedestrians and large groups . The new position of the display could encourage watching and mean that people are less con -cerned about lights turning red when they are halfway over the crossing.

- 4

-The green light can be on for short periods, because it is a start-signal only. Calls are cancelled one second after pedestrians leave the mat.

Observations were planned at two sites in Gouda and Heemstede, but prob-lems arose during installation of the mats in Gouda, leading to the can-cellation of the survey there. Chapter 2 describes the installation pro-cess and makes suggestions for improvements of infrared detectors and mats.

More than 1,000 pedestrians were observed. Their crossing and watching behaviour was noted in relation to the different phases, traffic flows and the presence of other pedestrians. The results of the video observation study are given in Chapter 3. Conclusions were drawn about their under-standing of PUSSYCATS.

Crossing on a red light, head movements, crossing between lines, conflicts and accepted gaps were observed as indicators of dangerous behaviour. Many factors, some clearly related to PUSSYCATS, were found to influence more or less dangerous behaviour. Waiting times and crossing speeds were observ-ed as indicators of the convenience of the system.

An estimate was made of the gain in time, as a measure of the system's efficiency. Vulnerable road users received special attention. The effect of the presence of other people was also determined.

Two hundred users of the crossing were interviewed, to obtain more infor-mation on their understanding of PUSSYCATS. They were asked to compare the old crossing with the new one, in terms of safety and convenience .

The results of this questionnaire survey can be found in Chapter 4.

In Chapter 5, conclusions are drawn about the efficiency, safety and con

CONTENTS

1. Introduction

2. Lessons from the installation of PUSSYCATS 2.1. Dutch real-life site Gouda

2.2. Dutch real-life site Heemstede

3. Observations of pedestrian behaviour

3.1. Introduction

3.2. Phase distribution: Signals timings 3.3. Vehicle and pedestrian flows

3.4. Composition of sample

3.5. Behaviour and understanding

3.6. Pedestrian use of crossing and risk taking 3.7. Pedestrian comfort and convenience

3.8. System safety and efficiency

3.9. Special groups and group behaviour 3.10. Social behaviour

4. Questionnaire survey

4.1. Introduction

4.2. Composition of the sample

4.3. Frequency of use: Questions 1-3

4.4. Observation and understanding of the new system: Questions 4-8 4.5. Waiting time perception: Questions 9A and 9B

4.6. Safety: Questions 10-15

4.7. Purpose of journey: Questions l6A 4.8. Publicity: Questions 17-18

4.9. Crossing behaviour: Questions 19-21

5. Conclusions

5.1. Operation and efficiency 5.2. Safety

5.3. Convenience

-1. INTRODUCTION

This report is the Dutch part of an international (French-British-Dutch) evaluation study of new pedestrian crossing facilities, summarized under the name 'PUSSYCATS' · It details the results of the installation process (Chapter 2), a video observation study (Chapter 3) and a questionnaire survey (Chapter 4). Conclusions on PUSSYCATS are drawn in Chapter 5.

The report follows the structure of the UK report as far as possible, to facilitate integration of the three reports. This basic principle also influences the presentation of data and tables to some extent.

The chapter on installation is an exception. The special Dutch situation, where peat soil made installation impossible at one site, is described in detail to explain why part of the study was cancelled. Observations and the survey had to be restricted to one site, Heemstede. However, the unsuccessful installation at Gouda proved very informative in terms of potential improvements to PUSSYCATS.

Many technical elements of PUSSYCATS are described in detail in other reports. This report, apart from aspects of installation, concentrates on the data of the field study on pedestrian behaviour and opinions.

A comparative study of the new and former situations was not accepted for DRIVE funding. However, many questions about PUSSYCATS can be answered by observing the behaviour of pedestrians using the crossing and by ques-tioning the users. Many of them are able to compare the old situation with the new.

Pro forma questionnaires and observations were developed in collaboration with British and French partners .

The Dutch study was made possible by support from the cities of Gouda and Heemstede, the Siemens Nederland and Nederland Haarlem manufacturing com -panies and the contractors VTN and Nettenbouw.

The video observations were prepared in cooperation with Messrs . J .G. Arnoldus and G.A. Varkevisser. The latter made the recordings and read out

- 8

-the tapes, in cooperation with Mr. P. Voorham. Mr. P.J.G. Verhoef pre-pared the ~tatistical analysis and analyzed the data in cooperation with Mr. F. Bijleveld. The survey was conducted in cooperation with Mrs. H. Hendriksen. Mr. U. Meier assisted in installation of the mats at Gouda. Mr. S. Oppe suggested many corrections and improvements.

2. LESSONS FROM THE INSTALLATION OF PUSSYCATS

2.1. Dutch real-life site Gouda

2.1.1. Description of the site

Gouda is a city in the 'green heart' of Holland, built on peaty soil. As a result, any structure which is not built on piles floats and, if heavy, slowly subsides.

Many elderly people live in the area. Shops are located at two sides of the intersection. The intersection is part of a complex system of two intersections, both controlled by the same computer. A map and diagram are provided in Appendix 1 and 2. A detection system for buses (Vetag) is integrated into the system.

The traffic flow is characterized by large streams of cars travelling in all directions, flows of cyclists, with separate detectors and traffic lights, some bus routes with computer-controlled priorities and a number of particularly elderly pedestrians. Towards the evening (6 p.m.) every-thing becomes very quiet.

2.1.2. Installation of the equipment

The City of Gouda and Siemens Nederland were kind enough to take respons-ibility for the installation activities and bore most of the costs. The installation work was carried out by VTN. Siemens supplied the controller and programming.

The installation of PUSSYCATS was part of a complete reconstruction of the two intersections and traffic control system. All four pedestrian cross

-ings at one intersection were equipped with PUSSYCATS. Siemens installed its own infrared detectors at two of the four crossings.

Infrared detectors

We had some general problems with the specifications and handbook:

- The nominal detection zone is only 3 meters wide. Dutch pedestrian crosssings are usually 4 or more meters wide. In many cases, crossings are also longer than the distance covered by two detectors. In Gouda, the crossing was 13.5 meters long.

- 10

-- The handbook only covered the case of MIPCs placed on a pole one meter beyond the crossing. We had to adapt this to a situation with a pole in

the middle.

Mats

The installation of the mats was unsuccessful . As delivery was six months late, Siemens Nederland was unable to prepare properly.

More importantly, some characteristics of the mats and the installation instructions were unsuitable for many Dutch situations.

- The organization of installation work may mean that wires have to be installed some days before the mats are placed. If this is the case, wires have to be cut through and reconnected afterwards, to ensure careful han-dling of the mats.

- Cables for many different purposes are often installed under footpaths. Local authorities then forbid concrete foundations, which would make the cables completely inaccessible.

- The wet soil means that the weight of the mat and foundation must equal the weight of the pavement, to ensure even subsidence. A concrete founda

-tion would have been too heavy in this case.

VTN sought an alternative for the concrete foundation. A thick iron plate was produced for the underside of the mat, but was not level enough. Treatment with polyester filler produced a completely level base, but it was still not possible to adjust the mat as specified.

In the meantime, two alternatives have been put forward for installation of mats in this type of situation. The first is to place the frame on a polystyrene block. Floating roads are built using this method. The other possibility is to develop a very light but strong frame, made of material used in aircraft construction. In any event, the manufacturer should be made responsible for developing a construction suitable for large parts of the Netherlands .

Some general installation problems were:

- It was fairly difficult to make a smooth connection between the mat and pavement.

- Paving stones in the Netherlands always measure 30

*

- Horizontal placement of the mat is a requirement which cannot be met, but also seems unnecessary.

Pedestrian signal heads

Siemens Nederland tried to control costs by using the old signal heads, positioned slightly lower and turned 180 degrees. However, as people stand-ing on the mat could not see these lights, a lower auxiliary signal head was added. Because of the position of the existing poles, signal heads could not be placed on the left, facing the oncoming traffic.

Controller

The controller was provided and programmed by Siemens Nederland. The computer programme gives priority to road users arriving first, with top priority to public transport buses. Road users cannot predict the order of the different stages or, therefore, the necessary waiting time and the following traffic direction. Only at night, when the road is empty, can they be sure that the green signal will appear immediately on detection. Parallel traffic may turn right while pedestrians are crossing.

Wait signal and other information

The old push button was removed, but the light behind the old button retained its function as a wait lamp, meaning that when someone stands on the mat, a call is set. The pedestrian is informed about the working of the new system by a pictogram and text (See Appendix 8) saying that pedes

-trians must (1) stand on the mat, (2) stay there until the green light ap-pears on their side, (3) cross between the lines and (4) remain attentive. This information has two functions: informing the pedestrian and avoiding liability problems. Positioning the signal head on the near side is ille-gal. However, as we could not find room for both the pictogram and text, we decided to show the text only.

The pedestrian green light is accompanied by a fast rattling sound, the red by a slow rattle. The fairly short green phase makes the audible signal necessary. When sound failed during installation, many pedestrians did not notice when the light turned green.

12

-2.1.3. Summary of results

The intersection at Gouda is appropriate for PUSSYCATS: a rather long crossing with many elderly pedestrians. The installation work provided some information about possible improvements of the infrared detectors and the manual. Coverage of a larger area would be preferable.

The installation of mats revealed serious problems with peat soils and cables under footpaths, characteristic of large parts of this country (50% of the western and northern Netherlands). Some suggestions for improvement of the construction and operation of the mats can be made; a better-fitting foundation and adjustment of the measurements of the mat to match the

dimensions of paving stones, for example.

2.2. Dutch real-life site Heemstede

Heemstede is a large town with a very complex and busy intersection near to the railway station. One crossing at this junction was equipped with PUSSYCATS .

2.2.1. Description of the site

Before the installation of PUSSYCATS, the crossing comprised two separate sections with a traffic island in the middle. The traffic lights in the middle were removed.

Two kinds of pedestrians can be distinguished: the two streams of commut-ers in the morning and afternoon, and people from 'over the bridge', who shop on the other side of the station during the day. Some of them carry their bicycles over the crossing.

The crossing was interesting, because many commuters violated the red light after pushing the button. During the day, red light violation was lower .

2.2.2. Installation of equipment

The Heemstede local authority, Nederland-Haarlem and Nettenbouw collabo-rated in the installation and covered the costs, with Nederland Haarlem contributing the major share. After the experiment, Nettenbouw will in-stall its own infrared detectors for further tests.

Infrared detectors

The special situation made it impossible to install infrared detectors on both sides of the crossing, as this would have required an extra pole. One detector was placed on one side of the crossing, the second on the island, facing the same way as the first. There was no evidence that this would hamper the functioning of the detectors. The length of the crossing was exactly the same as in Gouda: 13.5 meters.

Mats

No special installation problems were encountered. One detector was not sensitive enough, probably because it was not in a flat position. Touching the mat with the point of a shoe was enough to activate the system.

The following alterations in the functioning of the mat and/or detection board are suggested:

- The minimum weight of 20 kg per sensor is insufficient in many cases. If a pedestrian evenly divides his or her weight over two sensors, which rarely happens, he or she must weigh more than 40 kg to generate a call. Children of 5 weigh an average of 20 kg. Consequently, many young children will not be able to generate a green stage. We would prefer a minimum

detection threshold of 10 kg or less.

- If people move over the mat, detection can easily stop for a moment. The call disappears and the pedestrian returns to the back of the queue. A small adjustable delay before calls are cancelled would be desirable. Nederland Haarlem programmed this into the software, as in the UK, but this could conceivably be included in the detection board.

- The board inhibits activation of the sensors beyond a certain length of time. This adjustable period must be extended to the maximum waiting time for pedestrians. In Heemstede, we needed at least 180 seconds.

14

-Pedestrian signal heads

Nederland Haarlem developed a special signal head for placement on the near side of the crossing. The signal head should not be placed too high, to ensure that short pedestrians can also see the light. If it is placed too low, pedestrians sometimes block each other's view. Nederland Haarlem developed a signal head with light over 180 degrees, so that a group of pedestrians could see it from different directions.

If a second light is placed on an island in the centre of a crossing, people can become confused, thinking that the second light is their sig-nal. In that case, part of the 180 degree angle of the second signal head can be covered, so that it is no longer visible to pedestrians watching the first signal .

Because of the position of the existing poles, the signal heads could not be positioned to face the oncoming traffic. At one side, most pedestrians in fact chose the right side of the pedestrian display.

Controller

The controller was supplied and programmed by Nederland Haarlem·

The computer programme is characterized by first order control, giving priority to road users with the longest waiting time, together with a second order control, where the green light can be shown in the shadow of other, non-conflicting, signal groups.

Road users cannot predict the order of the different stages or, therefore, the waiting time and the following traffic direction· Only at night, when nobody else is present, can they be sure of an immediate green signal on detection · Without calls, all signals are on red.

During the pedestrian crossing phase no other traffic is allowed to pass .

The maximum waiting time is 180 seconds or more · The pedestrian green and flashing green lights are on for 4 and 3 seconds, respectively . The mini

-mal clearance time, if the infrared detector does not detect a pedestrian,

is 1 to 5 seconds, depending on the following stage. The maximum extra

clearance time, if pedestrians are detected by the infrared detectors , is 6 or 7 seconds, depending on the following stage. (See Scheme 1 and 2) .

Four different conflicting stages can precede and follow the pedestrian stage .

Phase:F A B C D F Time: 3 1-5 7 1-4 ... ---1---- --1---1---1---1---.. . 1---1--- ---.. . Time: 7-11 Phase: E

Scheme 1. Diagram of the pedestrian stage.

Phase

A. Flashing yellow other traffic B. Pedestrian clearance

Signal group: 3*

5 68 11

C. Pedestrian green plus flashing green D. Clearance

Signal group: 3

5 68 11

E. Maximum pedestrian extended Infrared clearance

Signal group: 3

5 68 11

F. Green for other signal groups

Duration (seconds): 3 4 1 3 5 7 2 4 2 1 9 11 9 7 indefinite

*

- 16

-Wait signal and other information

The wait lamp of the former push button equipment is used as the wait lamp. Only the push button was disconnected.

Both shields: pictogram and text are used.

The pedestrian green is accompanied by a fast rattle, the red by a slow rattle.

2.2.3. Summary of the results

The crossing is an interesting one for installation of the PUSSYCATS system. There are many red light violators and many use the push button without waiting for green light, causing loss in time.

The installation of the mats revealed that mats can be used in the Nether-lands, but that the operation could be improved. Higher sensitivity is desirable, such as a longer inhibition limit and an adjustable delay for cancelling calls. The signal head developed by Neder1and Haarlem is clear and strong.

3. OBSERVATIONS OF PEDESTRIAN BEHAVIOUR

3.1. Introduction

Multiple linear regression analyses were run, as in the British reports .

The results are presented in Appendix 9. The variables were chosen to explain the maximum variation, but most variables which make no signifi-cant contribution (~= .05) were left out. The interpretation of the results takes into account the fact that the choice of a variable some-times makes the effect of a correlated variable invisible. Appendix 9 describes the models and the interpretations. The results are presented in the main text.

Coefficients mentioned in the tables are raw coefficients. Their inter-pretation requires an idea of the possible values of the independent vari-able. For example, if crossing time is dependent on 'sex', a binary vari-able (0-1), and the value of the coefficient is 0.5, then the difference in crossing time for men and women will be 0.5 seconds. If crossing time depends on the length of the cycle (mean length: 286 seconds), and the value of the coefficient is 0.001, then 100 seconds of cycle length will

increase the crossing time by 0.1 seconds. Coefficients and standard error are rounded off to two decimal points unless the first decimal point is a zero. In this case, more decimal points are shown.

1. This chapter covers the findings of the video recordings made at the Heemstede test site. The new system became operational on 24 June 1991. The survey was conducted on 28, 29 and 31 August 1991, a Wednesday, Thursday and Saturday respectively. The pedestrians had only two months of experience with the new system.

2. Video recordings were made on four consecutive days in 1991~

Wednesday, August 28, 10.30-18.30 Thursday, August 29, 07.00-16.00 Friday, August 30, 09.30-18.30 Saturday, August 31, 09.00-15.00.

3. Signal controller information was not gathered in full. Only the times of the pedestrian phases in each cycle were recorded: the start and finish

18

-of the pedestrian green signal, the start and end -of the extension by the infrared detectors, and the start of the 'normal' red signal. A cycle is defined as the period from the start of one pedestrian green signal to the start of the next .

4 . The pro-forma used to collect data from the video was prepared in col

-laboration with the DRIVE partners and is given in Appendix 6.

5. Fourteen hours were selected for analysis (See Table 1).

The hours were selected to provide a sample of 1,000 pedestrians at all hours of a working day and during 'shopping time' on Saturday, to obtain as varied a picture as possible. All pedestrians were sampled during these hours. Data are available for 25 more pedestrians, arriving or leaving before or after the first or last pedestrian green observed on each day .

The Table shows working days and Saturdays, together with hours within a working day, to give an idea of the fluctuations during the day. The half hour between 16.00 and 16.30 is missing.

'N gr' 'ped' 'men, wom' '-20, -60 >60' 'fr.st, to st' 'vehic' 'motv' 'bicy'

the number of green pedestrian phases during the hour.

the number of pedestrians observed.

the number of male and female pedestrians. the age categories 0-20, 21-60, older than 60. coming from or going in the direction of the station, respectively.

number of vehicles, motor vehicles plus bicycles number of motor vehicles

number of cyclists

3.2 . Phase distribution: Signal timings

1. The appearance and order of different stages are not fixed, but are determined by demand, as is the length of the different phases and stages .

The pedestrian stage consists of the following phases (see Scheme 1 and Scheme 2): pedestrian clearance . green, flashing green, clearance and ex-tension of clearance time by the IR detector. The length of clearance time before the pedestrian stage ( 'IG ped') depends on the preceding stage. The length of clearance time after the pedestrian green depends on the follow

clear-07.23-08.00 13 60 48 12 5 55 0 5 55 466 420 46 91 13 7 4,1 48 10 4,8 2,2 178 13,7 8,1 2033 13 156 91,9 2211 08.00-09.00 26 98 65 33 21. 71 6 21 77 924 880 44 182 26 7 4,8 132 26 5,1 3,5 392 15,1 10,3 3413 26 131 89,7 3805 09.00-10.00 17 63 36 27 14. 30 19 23 40 478 463 15 119 17 7 3,5 99 17 5,8 2,9 269 15,8 7,8 3171 17 187 92,2 3440 10.00-11.00 20 67 27 40 7 33 27 34 33 353 341 12 140 20 7 3,8 126 20 6,3 3,4 326 16,3 8,8 3376 20 169 91,2 3702 11.00-12.00 11 40 11 29 8 25 7 15 25 384 379 5 77 11 7 2,1 66 11 6,0 1,8 176 16,0 4,7 3552 11 323 95,3 3728 12.00-13.00 15 65 28 37 20 39 6 23 42 339 333 6 105 15 7 3,1 80 15 5,3 2,3 230 15,3 6,7 3179 15 212 93,3 3409 13.00-14.00 19 76 32 44 10 58 8 46 30 413 402 11 133 19 7 3,7 100 18 5,6 2,8 290 15,3 8,1 3302 19 174 91,9 3592 14.00-15.00 17 60 32 28 8 31 21 37 23 451 435 16 119 17 7 3,2 107 17 6,3 2,9 277 16,3 7,4 3472 17 204 92,6 3749 15.00-16.00 15 50 28 22 10 32 8 28 22 523 486 37 105 15 7 3,0 83 14 5,9 2,4 233 15,5 6,6 3301 14 236 93,5 3531 W~t!D, A!'!Q, 211 16.30-17.30 17 84 49 35 11 59 14 45 39 802 774 28 119 17 7 3,5 100 17 5,9 2,9 270 15,9 7,9 3134 17 184 92,1 3404 17.30-18.30 18 106 62 44 21 80 5 73 33 863 805 58 126 18 7 3,3 86 17 5,1 2,3 266 14,8 7,0 3532 17 208 93,1 3795 Sill!.!. A!'!Q, 211 12.00-13.00 24 89 39 50 17 58 14 52 37 462 452 ₁₀₁₁₆₈ 24 7 4,6 153 24 6,4 4,2 393 16,4 10,9 3223 24 134 89,1 3616 13.00-14.00 14 70 37 33 13 47 10 32 38 470 448 22 98 14 7 2,8 111 14 7,9 3,1 251 17,9 7,1 3309 14 236 92,9 3560 14.00-15.00 18 73 30 43 12 48 13 32 41 538 527 11 126 18 7 3,4 120 18 6,7 3,3 300 16,7 8,1 3393 17 200 92,0 3690 ... Total 244 1001 524 477 177 666 158 466 535 7466 7145 321 1708 244 98 3,5 1411 238 5,9 2,9 3851 15,8 7,8 45390 241 188 92,2 49232 \0 Av. hour 17 72 37 34 13 48 11 33 38 533 510 23 122 17 3 101 17 3 275 8 3242 17 92 3517 av.cycle 4,1 2,1 2,0 0,7 2,7 0,6 1,9 2,2 30,6 29,3 1,3 7 7 5,9 5,9 15,8 15,8 188 188 202

N gr number of gl'8en phases vehic number of vehicles s gr seconds of green s pe seconds of pedestrian phase

ped number of pedestrians motv number of motorvehlcles N gr number of gl'8en phases av pe average pedestrian phase in seconds

men number of men bicy number of bicyclists av gr average green in seconds % pe % of pedestrian time

wom number of women % gr % of green time s 1'8 seconds of vehicle phase

-20 age: 0-20 s IR seconds of IR clearance N red number of vehicle phases

-60 age: 21-60 N IR number of IR clearance phases av 1'8 average vehicle phase in seconds

>60 age: >60 av IR average IR clearance in seconds % re % of vehicle time

fr.st coming from station % IR % of IR clearance time s tot seconds total

to st going to station

20

-ance time, caused by infrared detection, also varies according the follow-ing stage. Extended clearance time stops after one second of non-detection.

2. Table 1 shows the time in seconds, average time per cycle and percentage of time devoted to the different pedestrian phases.

After the clearance time (after pedestrian green), or after the extension of the green signal by the infrared detector, it takes some time before other traffic reaches the crossing. To make our data more comparable with the UK data, we added 3 seconds (known as 'clearance') to the pedestrian stage, an estimate of the mean duration before traffic reaches the cross-ing.

The clearance time before the pedestrian green varies. Again, we made an estimate of this clearance time (known as pedestrian intergreen: IG ped) of three seconds.

If there was no IR clearance, we restricted the pedestrian period after the green phase to three seconds. The clearance after pedestrian green is not mentioned in the remainder of this report, as it is either included in the IR clear or is considered to be included in the three seconds 'clear-ance'. This is not entirely justified, but is done to make data more com

-parable to the UK data.

The phases are referred to as:

'gr' green plus flashing green;

, IR' '. extension by IR detectors (IR clearance): the period from flashing green to the end of IR clearance;

'pe' total pedestrian time: gr + IR plus 3 seconds after the pedestrian phase (clearance);

, re' '. red, time for crossing vehicles, excluding 3 seconds intergreen after the pedestrian phase, but including 3 seconds intergreen before pedestrian green (IG ped) .

The prefix's' means seconds,' , N' is number of phases per hour; , av' i s

average number of seconds per phase " , %' is percentage of time devoted to the phase; 's tot' is the total number of seconds for which phases are measured during the hour.

The first hour starts at 07 ·23 a.m· and lasts only 2,211 seconds . The measurement starts at the first pedestrian green period after 7 a·m.

3. Observations were made during 244 cycles. As already mentioned, a cycle is defined a the period from the start of one pedestrian green to the start of the next. Sometimes the mat was used by a cyclist riding over the crossing when no pedestrians were present. These are not 'scored'. This occurred in 12 cycles. The other 232 cycles are real pedestrian cycles.

4. Sometimes, the green signal was immediately activated by a pedestrian or cyclist only touching the mat, sometimes when leaving the crossing. This is possible only if there are no calls by traffic participants else -where on the crossing, or if the pedestrian stage can be combined with another stage elsewhere on the crossing. Sometimes a green signal is fol-lowed immediately, or after some seconds of IR clearance, by another green signal, with no red signal in between.

5. The pedestrian time per hour (green, IR-clearance plus 3 seconds inter-green) as a percentage of total time per hour correlates with the number of pedestrians per hour (r ~ 0.61), not the number of vehicles crossing.

6. The vehicle time per hour bears almost no relation to the number of vehicles passing over the pedestrian crossing per hour, but is negatively related to number of pedestrians (r

=

7. The average pedestrian time per cycle was 15.8 seconds: 7 seconds green, 5.9 seconds IR clearance (not always present) and 3 seconds clearance. The average vehicle time was 188 seconds . The average cycle time was 202 seconds. This conflicts with the built-in maximum delay for the sensibility of the mat sensors, of 130 seconds. Pedestrians sometimes had to wait longer than 130 seconds. If they do not move to another sensor

in this period (each mat has 6 sensors) the call is cancelled.

The pedestrian time per cycle bears little or no relation to the number of pedestrians (0.13) or to the number of vehicles per cycle (0.11) . The pedestrian time is related to the number of pedestrians who arrive on green (0.22) or IR clearance (0.45) and to the number of pedestrians who start during IR clearance (0.47), but not to the number of pedestrians who start during green. Pedestrian time per cycle is related to the number of pedestrians per Cycle who arrive between the end of IR clearance of the

- 22

-preceding cycle and the end of IR clearance in the current cycle, and who start during green (0.29) or during IR clearance (0.73).

No relation was found between the mean crossing speed per cycle and the IR clearance time, nor between the crossing speed of the slowest pedestrian per cycle and the IR clearance time . Restriction of data to those pedes-trians who cross on green or during the IR clearance period does not make a difference.

8. The vehicle time per cycle is related to the number of vehicles per cycle (0.82) and to the number of pedestrians per cycle (0.59). Weak relationships exist between the pedestrian time per cycle and the vehicle time (0.17), and between pedestrian time and length of the cycle (0.18).

The cycle time is related to the number of vehicles (0.82) and the number of pedestrians (0.59) - not surprisingly, in view of the perfect correla-tion between cycle time and vehicle time (1.0).

9. Overall, two factors appear to influence pedestrian time. Firstly, the number of calls is important. When more pedestrians arrive, more pedes

-trian stages are realized, leading to a correlation of .74 between pedes-trians and number of cycles per hour. Secondly, the clearance time, and therefore the pedestrian time, is influenced by the presence of pedestrians on the crossing during this clearance period.

This leads to increasing correlations between pedestrian time per cycle and the number of pedestrians per cycle (.13), the number of pedestrians arriving on green (.22) or IR clearance (.45), the number starting during IR clearance (.47) and the number arriving between the end of IR clearance of the preceding cycle and the end of IR clearance in the present cycle who start during IR clearance ( .73).

Vehicle time is also influenced by two factors . Firstly, the number of pedestrian stages, caused by the number of pedestrians per hour, reduces the time assigned for vehicles, leading to a correlation of -.61 between vehicle time per hour and number of pedestrians. Secondly, vehicle time is influenced by the number of vehicles per cycle, producing a correlation of .82.

The only relevant differences in this respect between the old system and PUSSYCATS is the variable IR clearance and the cancellation of calls when people leave the mat. Pedestrian time is influenced by both factors. vehi-cle time almost exclusively by the number of calls realized and cancelled.

3.3. Vehicle and pedestrian flows

1. Table 1 also summarizes pedestrian and two-way vehicle counts per hour. Pedestrian flows are broken down by sex. age and direction (from and to the station). Vehicle flows are broken down by type of vehicle: all vehi-cles. bicycles included. motor vehicles and bicyvehi-cles.

2. Pedestrian rush hours are from 7.23 a.m. to 9.00 a.m. and from 4.30 p.m. to 6.30 p.m. (the first rush hour period only covers 37 minutes!). reflecting the fact that Heemstede is a commuter city and that the crossing leads to the railway station. The "direction" shows the same picture.

In the morning the flow is toward the station. in the afternoon in the opposite direction. The number of men exceeds the number of women during these hours. reflecting the uneven division of labour in the Netherlands. The age division is also as could be expected: the 21 to 60 age group is over-represented during these hours.

3. Traffic flow can be broken down by pedestrian stage and vehicle stage. The pedestrian phase is defined here by the start of the green signal and the end of IR clearance. See Table 2. The 30 vehicles during the pedestrian phase probably violated a red light .

Ped. stage Veh. stage Total Motor vehicles 5 7140 7145

Table 2. Vehicle counts.

Bicycles 25 296 321 Total vehicles 30 7436 7466

4. Traffic flow per hour is related to pedestrian flow. The correlation between vehicles per hour and pedestrians per hour is 0.74. The correla

-tion per cycle is .60 . The same patterns are found for motor vehicles (0.59) and bicycles (0 .46) .

24

-An explanation for these relations can be sought in the fact that, when the length of the cycle increases because more vehicles are detected, more pedestrians will arrive during this cycle, the average number of pedestri-ans per unity of time being equal. To assess this hypothesis, the correla-tion between the number of pedestrians and number of vehicles was control

-led for the length of the cycle. The result is a partial correlation of .21 which must be due to the fact that, as we have seen, at rush hours the pedestrian flow and the vehicle flow are higher than at other times. Very few motor vehicles cross during the pedestrian phase (green and IR clear-ance time): only five (out of a total of 7,145), and 25 bicycles (out of a total of 321).

5. Table 3 shows pedestrian counts for arrival phases, broken down by phase of starting and phase of finishing. 'IG ped' is added to 'Vehicle

Arrival Start Finish Count Percentage

Green Green Green 1 <1

Green Green IR clear 28 2.7

Green Green Clearance 16 1.6

Green Green Veh red 9 <1

Green IR clear IR clear 1 <1

Green IR clear Clearance 2 <1

Green IR clear Veh red 4 <1

Green Veh red Veh red 2 <1

IR clear IR clear IR clear 4 <1 IR clear IR clear Clearance 3 <1 IR clear IR clear Veh red 11 1.1 IR clear Clearance Veh red 4 <1 IR clear Veh red Veh red 2 <1

IR clear Green IR clear 5 <1

Clearance Clearance Veh red 2 <1 Clearance Veh red Veh red 7 <1

Clearance Green IR clear 5 <1

Clearance Green Clearance 1 <1

Veh red Veh red Veh red 528 51 .7

Veh red Veh red Green 33 3.2

Veh red Veh red IR clear 22 2.2

Veh red Green Green 5 <1

Veh red Green IR clear 263 25 ·7

Veh red Green Clearance 50 4.9

Veh red Green Veh red 13 1.3

Veh red IR clear IR clear 1 <1

red', because the extremely small numbers add no information. Vehicle red is defined as the period between three seconds after IR clearance and the next green signal.

The important flows are:

1. Arrival, starting and finishing on vehicle red (51.7%).

2. Arrival on vehicle red, starting on pedestrian green and finishing on IR clearance or clearance (30.6%).

The percentage of people arriving when the pedestrian signal was green was 8.3%. Green time covers only 3.5% of total time. Some people clearly anti-cipate the green phase.

The percentage of people arriving during the IR clearance period was 2.8%; IR clearance time is 2.9% of total time. People do not anticipate the green by starting in the IG ped period (three seconds before green): 2%. (IG ped is 1.5% of the total time).

3.4. Composition of sample

Table 4 summarizes the age/sex distribution of the sample.

Sex Male Female Total 0-10 count 15 10 25 % 1.5% 1.0% 2.5% 11-20 21-60 count % count 76 7.4% 383 81 7.9% 302 157 15.3% 685

Table 4. Age-sex distribution of sample.

60+ % count % 37.3% 65 6.3% 29.4% 94 9.2% 66.7% 159 15.5% Total count % 539 52.5% 487 47.5% 1026 100%

Few children were counted. The majority of pedestrians fell into the 21·60 age group.

168 (16.4%) of the 1,026 pedestrians were categorized under 'special circumstances' . Table 5 shows the observed special circumstances.

'Bicycle': many people walk in one direction to avoid crossing the street twice. In the other direction they use the bicycle. 82% of the 88 people walking with a bicycle were travelling towards the station.

No special circumstances O. Special circumstances 1. Blind

2. Wheelchair user

3. Walking difficulties

4. Young children walking

5. Pram/pushchair 6. Heavy load 7. Dog 8. Bicycle 9. Shopper 10. Other Total

Table 5. Special circumstances.

- 26 -Number 858 168 2 1 4 7 18 13 22 88 8 10 1026

3.5. Behaviour and understanding

Percentage 83.6 16.4 0.2 0.1 0.4 0.7 1.8 1.3 2.1 8.6 0.8 1.0 100

1. The push button and wait lamp are located in the same box. The push

button was not removed but was disconnected. Many people (23%) still use

the push button.

Women use the push button more often than men (31% vs · 16%), arrivers on

red more often then arrivers on IR clearance and green (25%, 6% and 2% respectively), and red crossers less often than green crossers (13% and

39%). Push button users have longer waiting times (r = 0.34), because of

green crossing. When other people are present on the same kerb on arrival, use of the push button is less (5% and 26%). The same is true for arrivals when the wait lamp is on, meaning that somebody is standing on the mat

(14% and 26%), or when other people are already crossing in the same

direction (9% and 25%). People who do not ~ediately stand on the mat, do

not use the push button (3% versus 47%). The push button can be used with

-out standing on the mat . People who use the push button watch the instruc

A relationship exists with age: the over-60s use the push button more often (40%) , 20 to 40-year-olds less often (20%).

2. Only those who arrive at the crossing when the wait lamp is off and the pedestrian signal on red ought to activate the system by standing on the mat. Table 6 shows the results. Half of the people (460) who should stand on the mat activate the system correctly. 14 more pedestrians stood on the mat after a short space of time (range 1 to 79 seconds; median: 5 seconds) . Of these, 11 arrived with the wait lamp off. No significant difference exists between people arriving with wait lamp on or off.

Wait lamp Standing on mat

yes no total

count % count % count %

On 96 44% 122 56% 218 100%

Off 364 49% 378 51% 742 100%

Total 460 48% 500 42% 960 100%

Table 6. Mat behaviour for those who arrive with pedestrian signal on red, broken down by wait lamp on or off on arrival.

3. The mat need only be used by pedestrians intending to cross on green. Table 7 distinguishes green and red crossers. 88% of pedestrians who arrive with the pedestrian signal on red and the wait lamp off, and who intend to cross on green, use the mat correctly. Red crossers use the mat much less frequently.

Crossing on Standing on mat

yes no total

count % count % count %

Green 208 88% 28 12% 236 100%

Red 156 31% 350 69% 506 100%

Total 364 49% 378 51% 742 100%

Table 7. Mat behaviour for those who arrive with the pedestrian signal on red and the wait lamp off, by green and red crossing.

- 28

-4. Seven more pedestrians arriving with the wait lamp off and intending to cross on green later stood on the mat, increasing the percentage to 90. The 10% not using the mat correctly were members of groups, or started together with other pedestrians. One can conclude that comprehension of the system was very good.

5. An advantage in time exists for other road users when pedestrians arriving with the signal on red and the wait lamp off use the mat, but leave before the light turns green. In this case, the call disappears unless other people stand on the mat in the meantime. An initial over-estimation of the advantage was produced by the 156 pedestrians (15% of all pedestrians) from Table 7 who arrived with the wait lamp off and stood on the mat, but then left before the light turned green. It is not known whether other pedestrians remained on the mat at that moment. We return to this point later.

17 pedestrians left the mat and stepped back on after 2 to 29 seconds.

3.6. Pedestrian use of crossing. risk taking

3.6.1. Red and green crossing

1. People can start crossing on: Red pedestrian signal:

- IR clearance

- clearance (3 seconds before veh red)

- veh red (the period where pedestrians have red, and traffic could pass) - IG ped (3 seconds before green)

Green pedestrian signal.

LG ped is hardly used and is therefore combined with veh red.

Start ing on green or IR clear is normally safe if people have enough tim'e to finish crossing before the end of IR clear, or at least of clearance·

2. Table 8 shows starting phases by finishing phases . 396 pedestrians (39%) behaved correctly by starting on green. 385 pedestrians (38%) can be considered to have crossed safely, by starting during the green or IR clear phase and finishing before veh red .

Starting on Finishing on

Green IR clear Clearance Veh red Total

count % count % count % count % count %

Green 6 1% 301 29% 67 7% 22 2% 396 39%

IR clear 6 1% 5 0% 15 1% 26 3%

Clearance 6 1% 6 1%

Veh red 33 3% 22 2% 539 53% 594 58%

Total 39 4% 329 32% 72 7% 582 57% 1022 100%

Table 8. Starting phase related to finishing phase.

3. Not all pedestrians starting on green finished crossing within the pedestrian time (green, IR clear and clearance). 17% (67/396) arrived during clearance. 5% (22/396) arrived during veh red.

The clearance, by our definition, does not guarantee that no traffic will be present. One can conclude that the maximum IR clearance should be

ex-tended by more than 3 seconds.

4 . The 'real' proportion of red light violation is the proportion of arrivers on red (617/960) who also start on red (during IR clear + clear-ance + veh red). This is 64.3% (See Table 9).

The proportion of 'real' green compliance is the proportion of arrivers on red who start on green (343/960). The percentage is 36%. Almost all of the 9 green arrivers who started on red, did so during the IR clearance time.

Arriving on Starting on

Green Red Total

count % count % count %

Green 54 86% 9 14% 63 100%

Red 343 36% 617 64% 960 100%

Total 397 39% 626 61% 1023 100%

30

-5. Table 10 shows some correlations between red light violation and flow variables. Red light violation per cycle is related to the number of

pedestrians (r = 0.91) per cycle, the number of vehicles (0 .64) per cycle, the length of the cycle (0.63) and the length of the vehicle phase (0.63) .

2 3 4 5 6 7 8 9 1. Start on green .09 .01 .40 .06 .03 .03 .09 0.01 2. Start on IR clear .03 .19 .01 .02 .02 .47 -0.04 3. Start on red .91 .64 .63 .63 .03 0.19 4. Number of pedestrians .60 .59 .59 .13 0.17 5. Vehicles .82 .82 .11 0.25 6. Length of cycle 1 .18 -0.09 7 . Length of red .17 -0.09 8. Length of IR clear -0.19 9. Rush hours

Table 10. Correlations between phase of start and flow variables, over 232 cycles.

A perfect correlation must be assumed between the length of the cycle and the length of the vehicle phase, with relationships between the number of pedestrians, number of vehicles and length of cycle. The question of

whether pedestrians are more willing to violate the red light when traffic flow is low can be answered by considering the correlation between 'start on red' and 'vehicles'. partialling out the effect of the length of cycle. as length of cycle is related to the number of vehicles and number of pedestrians. This partial correlation is 0 .27. We must conclude that red

light violation is higher when traffic flow is higher.

We have seen that a relationship exists between traffic flow and pedes -trian flow per hour (.74). A possible explanation is that during rush hours. pedestrians are less willing to wait for green because they are in a hurry. or because they are less obedient people .

6. Regression analyses are run to predict the number of red and green crossers per cycle . Then to predict red crossing of arrivers on red, and of arrivers on red who are confronted with the waitlamp switched on at a certain moment (Appendix 9, para. 1).

7. The number of red and green crossers per cycle is determined by the number of pedestrians per cycle (more pedestrians = more red and green

crossers), by rush hours, the 69 of the 232 cycles between 7 and 9 a.m.

and 4.30 and 6.30 p.m. (rush hours = more red and fewer green crossers),

by length of cycle (longer cycles

=

green crossers).

8. Important factors predicting red crossing by arrivers on red per pedes-trian are:

- Crossers towards the station cross on red more frequently (67% versus 61%), probably to catch a train.

- Age and sex are important factors. 11 to 60-year-olds are more willing to cross on red than younger and older pedestrians (See Table 11).

- Women are less willing to cross on red than men (See Table 12).

- People with dogs and bicycles cross less frequently on red (48% versus 66%), as do people in 'other special circumstances'.

- Group members cross less frequently on red then people walking alone (53% versus 68%).

- Red crossing increases when nobody is standing on the mat on arrival (51% versus 68%). Crossing on Age 0-10 count % 11-20 count % 21-60 count % 60+ count % Total count % Red Green Total 7 16 23 30% 106 70% 38 100% 144 74% 445 26% 197 100% 642 69% 59 31% 92 100% 151 Table 11. Red crossers of arrivers on red, by age.

Crossing Sex on Women Men count % count % Red 256 55% 361 73% Green 206 45% 137 28% Total 462 100% 498 100%

Table 12. Red crossers of arrivers on red, by sex.

39% 617 61% 343 100% 960 Total count % 617 343 64% 36% 960 100% 64% 36% 100%

32

-9. Pedestrians arriving on red and crossing on red are sometimes confront-ed with the wait lamp, because they themselves or other people stand on the mat. Sometimes they start before the wait lamp is switched on. Only in the first case could we expect to find a relationship between the length of the necessary waiting time and green/red crossing, because only in the case of a call for green will they receive information about the length of the necessary waiting time . The necessary waiting time is defined as the time between the last uncancelled call and the onset of green.

632 of the 960 red arrivers were confronted with the wait lamp on or after arrival. Table 13 presents the distribution of necessary waiting times, analysed by start phase (red or green). Pedestrians who have to wait longer are more inclined to cross on red.

Crossing on

Green Red

Necessary waiting time (seconds)

<11 11-30 31-60 61-100 101-160 161-260 >260 58 22 93 26 118 42 64 67 8 SO 3 39 6 58 Total 350 304

Table 13. Relation between red and green crossing and necessary waiting time.

10. It is important to note that neither model predicting crossing on red shows a negative relationship between red crossing and number of vehicles. It was generally expected that red light violation would be partly deter-mined by the amount of traffic: more traffic, would mean less red crossing .

We have found positive correlations between the number of vehicles and red crossing per cycle (r = .64) and per pedestrian (r = .27).

The traffic flow is probably so small that no negative relation can be found, but a positive relation would be unexpected. A possible explanation can be found in the behaviour of people during rush hours~ many red cross -ers, despite the large number of motor vehicles.

3.6.2. Head movements

1. Table 14 shows the number of pedestrians making head movements to right or left, in both directions, or making no head movements, before and while crossing.

Watch All pedestrians Right count % before crossing 89 9% while crossing 287 28% Red crossers before crossing 60 10% while crossing 196 31% Green crossers before crossing while crossing To the station 29 7% 91 23% before crossing 67 12% while crossing 215 39%

From the station before crossing while crossing 22 5% 72 15% Left count % 334 33% 79 8% 247 40% 46 7% 87 22% 33 8% 139 26% 21 4% 195 41% 58 12% Both count % 399 39% 383 37% 211 34% 259 41% 188 47% 124 31% 240 44% 194 36% 159 33% 189 39%

Table 14. Head movements before and while crossing.

Safe count % 733 72% 515 50% 458 74% 346 55% 275 69% 169 38% 379 70% 303 55% 354 74% 213 44% Not count % 204 20% 277 27% 108 17% 125 20% 96 24% 152 38% 100 18% 116 21% 104 22% 161 34%

2. Looking left before crossing means looking towards the oncoming

traf-fic. 'Safe' watching is therefore to the left and to both sides. 'Safe'

watching while crossing is less clear. If pedestrians have already at

least looked left before crossing, then looking right or in both direc

-tions can be described as 'safe'. Looking both ways is always safe . 'Safe'

while crossing is therefore 'both directions' plus part of 'the right

while crossing' figure, namely the proportion which looked 'left before' or 'both before' (See Table 14). The percentage of safe watching is 72

before crossing and 50 while crossing.

3. Watching is especially important for people who violate the red signal .

- 34 .

crossers are more likely to look left before crossing and right while crossing.

A relationship was found between head movements before and while crossing

(r = .32). This relationship remains in controls for red crossing

(part.cor = .30). Red crossers look more often, before and during cross·

ing, than green crossers. 15% of green crossers do not look at all, com

-pared with 8% of red crossers (see Table 15).

Before crossing Green crossers Watches Do not watch Total Red crossers Watches Do not watch Total All Eedestrians Watches Do not watch Total While crossing Watches count % 213 53% 35 9% 248 62% 445 71% 56 9% 501 80% 658 64% 91 9% 749 73% Do not watch count % 91 23% 61 15% 152 38% 73 12% 52 8% 125 20% 164 16% 113 11%

277

Table 15· Relation between crossing before and while

and red crossers.

Total count % 304 76% 96 24% 400 100% 518 83% 108 17% 626 100% 822 80% 204 20% 1026 100%

crossing, for green

4 · Pedestrians walking alone cannot trust the attentiveness of others and

are less likely to be distracted. Indeed, a weak relationship was found

between walking alone and head movements before (r

=

ing (.09). This is not only due to the fact that fewer group members are

red crossers. The first relationship remains significant in controls for

The light head should be positioned so that pedestrians watching the light are looking towards the oncoming traffic. In Heemstede this position could only be realized at the side of the station. Indeed, it was found that starters travelling towards the station looked left less (26%) than starters travelling from the station (41%) (See Table 14). The PUSSYCATS position of the display seems to increase watching in the direction of oncoming traffic.

5. Regression analyses were run to predict watching before and watching while crossing, for a number of variables (See Appendix 9, par. 2).

6. Overall, taking into account the fact that head movements are not necessary to detect approaching traffic and that the pedestrian green is conflict-free, 72% safe watching before crossing and 50% while crossing are not disappointing percentages, especially since red crossers look more closely than green crossers.

Members of groups are less likely to look. They probably trust each other. It is generally accepted that crossing in groups is safer than crossing alone. There are two ways of crossing in groups: as a member of a group, or together with other - unknown - people. Unlike group members, people

'crossing together' look more closely before crossing. However, the effects of group membership, together with other factors such as 'number of vehicles' and 'age', are difficult to isolate from the effect of cross-ing on red, and vice versa.

An important aspect of PUSSYCATS is the position of the light-head facing the oncoming traffic. This seems to improve closer watching before cross

-ing.

3.6.3. Crossing between lines

1. The crossing consists of two sections with an island in the middle .

Table 16 shows the number of pedestrians crossing between the lines of the first and second sections.

All pedestrians (1026) Green crossers (403) Red crossers (623) From station (480) Towards station (546) All pedestrians (1026) 36 -First part count % 623 61% 313 78% 310 50% 322 67% 301 55% Near station 646 63% Second part count % 628 61% 290 73% 338 54% 304 63% 324 59% Opposite station 605 59%

Table 16. Crossing between lines on the first and second section of the crossing and on the parts nearest and furthest from the station.

2. Crossing between the lines of the first section is more important than on the second, because the infrared detector stops functioning after 1 second if no pedestrian is detected. However Table 16 shows no difference between the first and second sections.

3. The layout of the crossing is such that the section furthest from the

station does not meet the street corner. The distance to the corner is 7 meters. People could be expected to avoid that part of the crossing in

particular. However, the difference is very small (see Table 16 below) .

When crossing from the station, people cross more often between the lines

of the first section (67%) than when crossing to the station (55%) . No

difference is found for the second section.

4. We might expect people who cross between the lines of the first section

to also cross between the lines of the second. Table 17 shows that this

often is the case (r

=

(r = .45) than for red crossers (r - .23) . As we might expect, this was

those travelling to the station (to station: r = .41; from station:

r = .24). The same is shown by Table 17 If there were no relationship

between crossing between the lines on the first and second section, we could expect the percentages shown in the table under 'ind.exp.%'.

Another point is that the smaller group of red crossers who cross between the lines do it more consistently on both sections. This is reflected in the difference between '%' and 'ind.exp.%', which is larger for red cross-ers than for green crosscross-ers.

First part Between lines Outside lines Green crossers between lines outside lines Red crossers between lines outside lines From station between lines outside lines Towards station between lines outside lines Second part Between lines count % ind.exp.% 463 165 260 30 203 135 230 74 233 91 74% 41% 83% 35% 66% 43% 71% 47% 77% 37% 37% 57% 27% 42% 23% count 160 238 53 57 107 181 92 84 68 154

Outs ide lines % ind.exp.% 26% 59% 17% 66% 35% 57% 29% 53% 23% 33% 15% 6% 23% 12% 18%

Table 17. Crossing between the lines, first and second section, from and

towards station, and expected percentages if crossing between lines on

first and second section were not related.

5. Looking and crossing between the lines are expressions of care. A rela-tionship can be expected between looking before crossing and crossing

38

-between the lines on the first section· This proves to be the case for red

crossers (r = .18), but not for green crossers (r = .07). Table 18 presents

the relationship between looking and crossing between the lines for the two groups. Green crossers are more likely to look and cross between the lines than red crossers. Green crossers also cross between the lines more often, but red crossers more often look before crossing. Looking is more important for red crossers, because of potential conflicts. Crossing be-tween the lines is more important for green crossers, because of the IR

detectors . We can conclude that a relationship exists between the logic of

the system and the behaviour of green and red crossers.

Between lines Outside lines Total

count % count % count %

Green crossers watches before 243 61% 61 15% 304 76% do not watch 70 18% 26 7% 96 24% Total 313 78% 87 22% 400 100% Red crossers watches before 278 44% 240 38% 518 83% do not watch 32 5% 76 12% 108 17% Total 310 50% 316 51% 626 100%

Table 18. Looking before crossing and crossing between the lines on the

first section.

6. Regression analyses were run to predict crossing between the lines of

the first and second sections (Appendix 9, para. 3) .

7. Overall, general factors and factors related to PUSSYCATS can be

distinguished. In general , the layout of the crossing does not regulate

crossing between the lines . Crossing between the lines of the first se

C-tion makes crossing between the lines of the second considerably more

likely. Green crossers cross between the lines more often than red cross

In relation to PUSSYCATS, we found that standing on the mat almost auto-matically led to crossing between the lines, which is necessary to trigger

the infrared detector, and that green crossers very often continue cross-ing between the lines, which is necessary to keep the infrared detector functioning.

3.6.4. Potential conflict between pedestrians and vehicles

1. In 36 cases, pedestrians and vehicles were observed in conflict, inclu-ding 34 different pedestrians (3.3% of all pedestrians). The analysis is confined to 34 conflicts in 14 hours of observation. Two pedestrians had two conflicts while crossing. A distinction can be made between situations where pedestrians have to rush forward to avoid a vehicle (27: 75%) and where they have to stop to avoid a passing vehicle (9: 25%). A conflict is defined as an encounter where at least one party has to change direction or speed to avoid a collision. In all conflicts, the pedestrians, at least, had to change pace. The danger of the conflicts could not be assessed, due

to lack of time.

All but four cases involved pedestrians who violated the red light. 4.8% of all red crossers come into conflict, compared with 1% of green cross-ers. We can conclude that crossing on green has a lower risk potential than violating red.

The four green crossers finished crossing in the IR clearance period. The other pedestrians in conflict situations finished on red.

2. The location of the conflicts is interesting. Most occur on the second part of the crossing, particularly on the section nearest the station.

(See Table 19). This section is clearly the section most dangerous. Pedes-trians face traffic from three directions, with different starting moments.

Part of crossing First part Second part Total From station 3 5 8 3 23 26 Towards station 6 28 34

Table 19. Points of conflict and section of crossing.