The UK Intelligent Speed Adaptation (ISA) project produced a rich database with high-resolution dataon driver behaviour covering a comprehensive range of road environment. The field trials provided vitalinformation on driver behaviour in the presence of ISA. The purpose of this paper is to exploit the infor-mation gathered in the field trials to predict the impacts of various forms of ISA and to assess whetherISA is viable in terms of benefit-to-cost ratio. ISA is predicted to save up to 33% of accidents on urbanroads, and to reduce CO2 emissions by up to 5.8% on 70 mph roads. In order to investigate the long-termimpacts of ISA, two hypothetical deployment scenarios were envisaged covering a 60-year appraisalperiod. The results indicate that ISA could deliver a very healthy benefit-to-cost ratio, ranging from 3.4to 7.4, depending on the deployment scenarios. Under both deployment scenarios, ISA has recovered its
ost benefit analysis implementation costs in less than 15 years. It can be concluded that implementation of ISA is clearlyjustified from a social cost and benefit perspective. Of the two deployment scenarios, the Market Drivenone is substantially outperformed by the Authority Driven one. The benefits of ISA on fuel saving andemission reduction are real but not substantial, in comparison with the benefits on accident reduction;up to 98% of benefits are attributable to accident savings. Indeed, ISA is predicted to lead to a savings of30% in fatal crashes and 25% in serious crashes over the 60-year period modelled.
. Introduction
The literature on the relationship between both individual vehi-le speed and traffic speed on the one hand and crash risk on thether is very extensive. A useful review of the literature on the sub-ect was performed by Aarts and van Schagen (2006), while Fincht al. (1994) examined the relationship between changes in speednd speed limits and the consequent changes in accident numbers.hey concluded that there was substantial support for a relation-hip in which a 1 km/h change in traffic speed on a road led to a 3%hange in the number of accidents on that road.
Intelligent Speed Adaptation (ISA) is a system that brings feed-ack about speeding behaviour into the vehicle. It may merely warnhe driver about speeding or it may intervene to prevent speeding.
number of studies of driver behaviour with various kinds of ISAystems have been reported (e.g. Duynstee et al., 2001; Lahrmannt al., 2001; Hjälmdahl and Várhelyi, 2004; Ehrlich et al., 2006).
However, there has been relatively little analysis of the overallmpacts that deployment of ISA would bring to the quality of theraffic as a whole. Carsten and Tate (2005) predicted the accident
reduction from a wholesale deployment of ISA, but the predictionswere based not on speeds observed from real-world experiencewith ISA speed but rather on a literature review that mainly usedstudies of compliance with speed limit advice from the roadside.The high-resolution data collected from the ISA-UK project pro-vides a rich database for establishing a comprehensive picture ofthe potential impact of ISA on speed reduction (Lai et al., 2008).The purpose of this paper is to explore the information gathered inthe field trials in order to predict the impacts of various forms ofIntelligent Speed Adaptation; i.e. what does such a change in speeddistribution imply in terms of system effects? Accident reductionis the most widely accepted measures for road safety, and it is wellestablished that there is a positive association between speed andaccident occurrence as well as severity (Nilsson, 1982; Elvik et al.,2004). Hence, how many accidents could ISA potentially save? Inaddition, ISA curtails excessive speeds towards the high end of thespeed distribution, and hence the implications for fuel consumptionand carbon dioxide emissions are also examined.
Three variants of ISA have been considered within the contextof this paper:
• Advisory ISA which informs the driver of the speed limit and warnsthe driver when the limit is being exceeded.
Baseline (no ISA) Speed distributions from Phase 1 of the ISA-UKdata (70,457 miles)
Advisory ISA system Apply the proportional reductions of observedfrom the LAVIA results (Ehrlich et al., 2006) tothe speed distributions obtained from theISA-UK data.
Voluntary ISA system All data from Phase 2 of the ISA-UK trials(211,198 miles)
Mandatory ISA Data from the non-overridden component of
4 F. Lai et al. / Accident Analysi
Voluntary ISA in which the information on speed limit is linked tothe vehicle’s engine management system and perhaps addition-ally to the braking system—the system comes on with the vehicleignition, but it may be overridden by the driver at will.Mandatory ISA which works like Voluntary ISA, but without theoption to override.
Each of this has been considered in terms of an overall ISA sys-em capability that covers all fixed speed limits. In addition, thereould be many different versions of the future; two example sce-arios were developed for consideration of the benefit from ISAeployment:
A Market Driven scenario in which users choose to have ISAbecause they want it, andAn Authority Driven scenario in which adoption of ISA, particu-larly ‘stronger’ forms of ISA, is initially encouraged and eventuallyrequired.
Following the monetisation of the potential benefits, the costsf implementing ISA was also analysed allowing the benefit-to-ost ratio (BCR) to be calculated. The BCR can be seen as indicatinghether there is a case for proceeding with ISA implementation on
he grounds that the total benefits that accumulate over the periodlearly outweigh the costs of investing in ISA.
This paper is organised into seven sections. The next sectionescribes the source of data, followed by the analysis of the impactsf ISA on accident reduction, fuel consumption and emission reduc-ion. The long-term impacts of ISA are then investigated, followedy discussions.
. Source of data
This study primarily made use of the data collected from the ISA-K field trials, which used a fleet of 20 identical vehicles. Seventy-ine drivers participated in the trials; each had the ISA car for aix-month period. The field trials had three distinctive phases asollows:
Phase 1: one month, when the ISA functionality was switched off.Phase 2: four months, when the ISA system was enabled. The ISAfunctionality was switched on upon engine ignition and the vehi-cle’s top speed was constrained at the prevailing speed limits. Theparticipants however could turn the ISA system off, temporarily.The speed limitation would be resumed when the driver switchedthe system back on voluntarily, when the vehicle speed droppedbelow the speed limit (e.g. in response to the lead vehicle slowingdown), or when the vehicle encountered a new speed limit.Phase 3: one month, when the ISA functionality was againswitched off.
The field trials collected 429,487 miles driving data inotal, of which 354,592 miles were with a known speed limitttached—these provided the basis for this analysis on the threeariants of ISA.
Data for Voluntary ISA system were from Phase 2 of the ISA-K trial, which allowed drivers to override the ISA system. Data forandatory ISA system were also derived from Phase 2; driving data
ecorded when ISA was overridden were excluded (i.e. hence theemaining data from Phase 2 represented what the vehicle speedould be like if the system was non-overridable). An Advisory ISA
ystem was not tested in the ISA-UK trials. However, the FrenchSA project, LAVIA (Ehrlich et al., 2006), tested advisory and volun-ary systems. Since the design and functionality were very similaretween the UK and the French ISA systems, data for advisory ISA
Phase 2 of the ISA-UK trials (191,582 miles)
for this analysis were modelled based on recorded speeds with vol-untary ISA from ISA-UK adjusted by the relative impact on speedchoice of an advisory ISA as compared to a voluntary ISA obtainedfrom the results of the LAVIA project. These data sources are sum-marised in Table 1.
3. The effect of ISA on accident reduction
ISA introduces a particular way of affecting speed choice andhence risk. ISA has been evaluated in a number of trials across theworld but none of these trials, including the ISA-UK trials, havebeen of sufficient size to provide reliable empirical information onactual crash involvement. To arrive at a conclusion about the safetyefficacy of ISA, it is necessary to apply models relating speed tocrash risk. The observed changes in speed choice (or the estimatedchanges in speed choice for alternative variants of ISA) can then beused to predict changes in crash or injury risk.
3.1. Application of speed–crash relationships to the ISA-UK data
Although there is considerable literature on the impact of speedon accident risk (e.g. Elvik et al., 2004, for a review and meta-analysis), not all the models generated from empirical data areequally applicable to the case of the introduction of ISA. The ISA-UKproject reviewed relevant models (Tate and Carsten, 2008). Therewere three major considerations in model selection:
• It is preferable to use models developed based on case-controlleddata rather than cross-sectional data. This is because roads withdifferent speed profiles tend to vary in quality and traffic con-ditions, and this may not be fully accounted for in the modelparameters.
• Whether the models cater to changes in the shape of the speeddistribution. ISA, in particular the intervening forms of ISA, rad-ically alters the shape of the speed distribution by curtailingspeeds in excess of the limit. This renders those models that con-sider only mean speed imperfect for ISA application on theoreticalgrounds.
• It is preferable to use models developed based on UK data tominimise any locality effect.
No single model fits in all of the requirements. A set of mod-els developed based on Australian data (Kloeden et al., 2001, 2002)satisfy the first two criteria, and were considered to be the mostsuitable candidates for this analysis. A risk curve is proposed byKloeden et al. (2001, 2002). The application of the risk curve to the
ISA data is illustrated in Fig. 1. The areas under the risk profiles inthe lower graph represent the expected relative crash risk. By com-paring the expected risk between the baseline condition and the ISA
F. Lai et al. / Accident Analysis and Prevention 48 (2012) 63– 72 65
Baseline ISA
Area under curve = 0.71
Area under curv e = 0.33
0%2%4%6%8%
10%12%14%16%18%20%
0 10 20 30 40 50 60 70 80 90 100
Prop
or�
on o
f jo
urne
y
Vehicle speed (mph )
0%2%4%6%8%
10%12%14%16%18%20%
0 10 20 30 40 50 60 70 80 90 100
Prop
or�
on o
f jo
urne
y
Vehicle speed (mph )
02468
101214161820
0 10 20 30 40 50 60 70 80 90 100
Rela
�ve
Ris
k
Vehicle speed (mph)
02468
101214161820
0 10 20 30 40 50 60 70 80 90 100Re
la�
ve R
isk
Vehicle speed (mph)
00.010.020.030.040.050.060.070.080.09
0.1
0 10 20 30 40 50 60 70 80 90 100
Risk
pro
file
(%*r
isk)
Vehicle spee d (mph)
00.010.020.030.040.050.060.070.080.09
0.1
0 10 20 30 40 50 60 70 80 90 100
Risk
pro
file
(%*r
isk)
Vehicle speed (mph )
rofiles
ct
Hb
cwtgKt
C
wirf
n
Fig. 1. Comparison of risk p
ondition, a crash reduction ratio can be obtained. In this example,he accident reduction ratio would be:
0.330.71
= 0.46
ence, the expected number of crashes in the presence of ISA woulde 46% of the crashes that would occur in the absence of ISA.
Although the Kloeden models purport to represent all injuryrashes, the crashes included in the studies were in fact those tohich an ambulance was called. As a result the crash set is bet-
er represented as constituting the killed or seriously injured crashroup. To overcome this bias, the crash reductions predicted by theloeden models have been adjusted. The adjustment is based on
he power model (Elvik et al., 2004; Table 14) as follows:
all injury = Xb/c
here Call injury is the expected reduction in all injury crashes; X
s the crash risk reduction predicted by Kloeden; b is the powerelationship for all injury crashes 2.78; c is the power relationshipor fatal and serious injury crashes 3.41.
Applying this adjustment to the example of Fig. 1, the expectedumber of injury crashes in the presence of ISA is adjusted as
between baseline and ISA.
follows:
0.46
(2.783.41
)= 0.54
Hence, after the adjustment, the expected number of crashes in thepresence of ISA would be 54% of the crashes that would occur inthe absence of ISA; this is a more conservative estimation than theapplication of the Kloeden model without adjustment.
There are two sets of Kloeden models, based on data collectedfrom urban roads with a 60 km/h speed limit (Kloeden et al., 2002)and rural roads with 80–110 km/h speed limits (Kloeden et al.,2001) respectively. Speed profiles obtained from the ISA-UK trialswere plotted according to speed limit, road type, and ISA variants.The shape of the speed profiles was examined. Where a spiky shapewas present, the Kloeden models were preferred. Where a moreconventional bell shape was present (e.g. Advisory ISA on mostroads) or the roads are out of the applicable range of the Kloedenmodels, alternative models were considered.
For urban environments, the U2 model (Taylor et al., 2000) hasthe advantage of targeting at the higher end of the speed distribu-tion. There were theoretical concerns over the applicability of theU2 model to ISA, because it would predict zero crashes in the case
6 s and
otHaayEd
ddmctooCCcnfibIwdp
Ta0o
tmnTowrIism(s
nImvbFc“osbccvc
3
b
6 F. Lai et al. / Accident Analysi
f non-existence of speeding (e.g. a perfect system configuration ashe Mandatory ISA at 100% penetration across the entire network).owever, the ISA system developed for the ISA-UK trials did allow
small amount of over speed limit before intervention would bectivated, which renders the U2 model being applicable to this anal-sis. For 70 mph roads (motorways and rural dual-carriageways),lvik’s power model was used because the Kloeden rural model waseveloped based on single carriageways.
It is recognised that using a combination of different models forifferent road types is likely to result in different errors applying toifferent situations. However, the alternative—applying the sameodel throughout—is considered less desirable, given the dramatic
hanges in speed distribution generated by Voluntary and Manda-ory ISA, and the inability of some models to take adequate accountf the impact of ISA on higher-speed vehicles. Sensitivity testsn different combinations of models were carried out (Tate andarsten, 2008). The final selection of models is reported in Table 2.ells filled with a cross indicate the absence of valid data for theombination of speed limit and road type; for example, there areo data applicable to motorways with a 20 mph speed limit. Cellslled with a dash indicate that insufficient data from these com-inations of road types and speed limits were recorded during theSA trial, which did not warrant their inclusion in the analysis. It is
orth noting that these cells representing only a handful of acci-ents on UK roads and hence are deemed to be negligible for theurpose of this analysis.
The resulted accident reduction potentials are reported inable 3. The numbers represent a ratio of the accident risk betweenn ISA variant and the baseline (i.e. no ISA). For example, a ratio of.8 indicates that the ISA system reduces the likelihood of accidentccurrence by 20%.
Mandatory ISA achieves the greatest potential of accident reduc-ion across all combinations of road types. The differences between
andatory ISA and the other two variants of ISA are most promi-ent on urban roads, where a large proportion of accidents occur.he effect of voluntary ISA obviously is subject to the amount ofverriding. The results from the ISA-UK trials indicate that ISAas most often overridden on urban A roads and motorways, thus
educing the additional impact of a voluntary ISA over an advisorySA as shown in Table 3. The effect of Advisory ISA is generally min-mal. Overall it can be seen that the impact of advisory ISA is quitemall. But the impact of intervening forms of ISA—voluntary andandatory—is predicted to be quite large, particularly on urban
30 and 40 mph) roads. This means that ISA has a large potential toave accidents involving vulnerable road users.
The fact that that a vehicle is equipped with ISA will notecessarily affect its risk at the accident level. For example, the
SA-equipped vehicle might be the vehicle setting off from theinor road in a two-vehicle crash at an intersection. For single-
ehicle accidents, the issue is relatively straightforward: risk wille affected by the proportion of ISA-equipped vehicles on the road.or two-vehicle crashes, it can be assumed that the proportion ofrashes affected by ISA will be equal to the proportion of ISA in theat fault” vehicles. This again will at least be equal to the proportionf ISA vehicles overall. It is actually a conservative assumption,ince in head-on collisions, for example, both participants coulde affected by ISA, and even rear-end collisions at high speedsould be affected by both vehicles having ISA. For multiple-vehiclerashes, it is again assumed that risk will be affected by at least oneehicle having ISA, and once again this assumption is somewhatonservative.
.2. Overall safety impact of ISA
The accident reduction potential of ISA reported in Table 3 isased on the vehicle speed profiles derived from the ISA-UK trials
Prevention 48 (2012) 63– 72
(i.e. 79 drivers and 20 vehicles). In order to examine the effect ofISA on vehicle speeds along an increase in system penetration, theAIMSUM micro simulation package was used to predict speed pro-files along hypothetic penetration levels of ISA at 20%, 40%, 60%,80%, and 100% across a range of road networks. The hypothesishere is that, at higher levels of penetration, ISA-equipped vehicleswould have a knock-on effect on the speed of vehicles followingthem in the traffic stream. Such effects were observed in some net-works, particularly the motorway network where trucks limited to60 mph tended to act as moving obstacles. But overall, there was nosubstantial confirmation of anything apart from a linear increase inISA impact with increasing penetration. The assumption of a linearrelationship is rather conservative, but has the advantage of avoid-ing over-estimation of the accident reduction potential at higherlevels of penetration that are still below 100%.
The accident analysis used five years worth of accident records(between 2002 and 2006) held in the UK’s national database. Theaccident reduction rates were applied to corresponding numbersof recorded accidents according to road types and speed limits topredict the proportion of accidents that would be saved by ISA.As shown in Table 4, the voluntary and mandatory variants of ISAare estimated to be considerably more effective than advisory ISA.Effectiveness goes up with penetration level. At 100% penetration,voluntary ISA would reduce the number of injury accidents by12% and mandatory ISA by 29%. On motorways, as compared toall roads, ISA is predicted to have a smaller but still considerablesafety impact. The impact of ISA is greatest on 30 mph roads, whichis logical because many of these crashes involve a collision with apedestrian where vehicle speed strongly affects the risk of severeinjury or fatality. On such roads ISA has the potential to save up toone-third of injury accidents.
4. The long-term impact of ISA
4.1. Deployment scenarios
The results in Table 4 demonstrate how ISA would reduce acci-dent occurrence at various levels of penetration. However, ISAcannot be introduced overnight, and a considerable time mightelapse before ISA reached 100% penetration. During such a period,traffic volumes would almost certainly grow, and safety would gen-erally improve because of measures other than ISA. Both trafficgrowth and the tendency for accident risk to reduce over time weretherefore taken into account in the analysis by applying the officiallong-term forecasts (Carsten et al., 2008).
The long-term impact of ISA was assessed based on a 60-yearperiod, as recommended for appraisal of investment in transportin the UK (Department for Transport, 2003), with ISA deploymentassumed to commence in 2010. The number of accidents in the baseyear was derived from 5 years of historical accident data between2002 and 2006 taken from the UK’s national accident database,STATS19, and rolled forward to 2010. In addition, there are numer-ous ways for ISA to be deployed into the vehicle fleet. For thepurpose of this analysis, two versions of deployment scenarios wereenvisaged.
• A Market Driven scenario in which users choose to have ISAbecause they want it. This scenario assumes a high proportionof non-intervening type of ISA in the fleet; e.g. advisory ISA, mostlikely to be wide spread as an add-on to OEM or after-marketsatellite navigation systems. Some drivers and fleet managers
would choose to have voluntary ISA (i.e. intervening but over-ridable) due to its safety benefits. The penetration of voluntaryISA is expected to be rather slow. The Market Driven scenario isillustrated in the left graph in Fig. 2.
F. Lai et al. / Accident Analysis and Prevention 48 (2012) 63– 72 67
Table 2The models applied for the accident reduction analysis.
Road type ISA variant Speed limit
20 mph 30 mph 40 mph 50 mph 60 mph 70 mph
Unclassified roads Advisory U2 U2 U2 P P –Voluntary U2 KU KU P P –Mandatory U2 KU KU P P –
B roads Advisory U2 U2 U2 KR KR –Voluntary U2 KU KU KR KR –Mandatory U2 KU KU KR KR –
A roads Advisory – U2 U2 KR KR PVoluntary – KU KU KR KR PMandatory – KU KU KR KR P
Motorways Advisory X – U2 KR – PVoluntary X – KU KR – PMandatory X – KU KR – P
Note: U2: U2 model (Taylor et al., 2000); KR: Kloeden rural model (Kloeden et al., 2001); KU: Kloeden urban model (Kloeden et al., 2002); P: Elvik power model (Elvik et al.,2004).
Table 3Predicted injury accidents reduction ratios for cars by ISA variant and road type.
Fig. 2. Penetration of ISA under different deployment scenarios.
68 F. Lai et al. / Accident Analysis and Prevention 48 (2012) 63– 72
Market Dr iven scenario Authori ty Dr iven scenarioFatal injuries accidents Fatal injuries accid ent s
Serious in juries accidents Serious in juries accide nts
Slight in juries accidents Slight in juries accidents
0
1
2
3
2010 2020 2030 2040 2050 2060 2070
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
0
1
2
3
2010 2020 2030 2040 2050 2060 2070
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
0
5
10
15
20
25
30
2010 2020 2030 2040 2050 2060 2070
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
0
5
10
15
20
25
30
2010 2020 2030 2040 2050 2060 2070
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
0
50
100
150
200
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
0
50
100
150
200
No
ofac
cide
nts
(1,0
00)
Without ISAWith ISA
ccurre
•
4
bdrtArni
2010 2020 2030 2040 2050 2060 2070
Fig. 3. Predicted accident o
An Authority Driven scenario in which adoption of ISA is initiallyencouraged and eventually required, with greater emphasis tovoluntary ISA and to fitment in newly registered vehicles. Retrofitof ISA would also be implemented at a faster pace to acceleratetake-up, especially in the vehicle fleets within the public sector.Upon full penetration of voluntary ISA, mandatory ISA (i.e. non-overridable) is implemented to maximise the safety potential ofISA. The Authority Driven scenario is illustrated in the right graphin Fig. 2.
.2. The long-term impacts of ISA on accident reduction
The predicted accident occurrence is illustrated in Fig. 3. Theaseline condition (i.e. without ISA) shows that accident occurrenceecreases in the long-term but ISA further reduces the accidentate. Under the Authority Driven scenario ISA delivers substan-ially greater safety benefits than under the Market Driven scenario.
cross both scenarios, ISA consistently delivers greater benefits ineducing more severe accidents. Table 5 summarises the predictedumber of accidents over the 60-year appraisal period. The Author-
ty Driven scenario is predicted to save 30% of fatal accidents and
2010 2020 2030 2040 2050 2060 2070
nce with and without ISA.
25% of serious accidents, and the Market Driven scenario wouldsave 13% of fatal accidents and 8% of serious crashes. Overall, 16%and 5% of road accidents would be saved by ISA under the AuthorityDriven and the Market Driven scenarios respectively.
4.3. The long-term impact of ISA on fuel consumption and carbondioxide emissions
ISA curtails excessive speeds and hence would reduce fuelconsumption and carbon dioxide emissions. In addition to safetybenefits, the impact of ISA on environmental benefits was alsoinvestigated using the data collected from the UK ISA project.This comprehensive data set provides an opportunity to estimatevehicle emissions of CO2 in a very detailed way by capturingthe actual transient nature of real driving conditions (e.g. vehicleacceleration). A flexible statistical modelling technique known asGeneralized Additive Modelling (GAM) (Hastie and Tibshirani, 1990)
was adopted to develop emissions models at a 1-Hz time resolutionwhich were then used to examine the effect of ISA on carbon dioxideemissions (Carslaw et al., 2010). The impact of ISA on CO2 emissionsis generally more significantly on high speed roads; voluntary ISA
F. Lai et al. / Accident Analysis and Prevention 48 (2012) 63– 72 69
Table 5Predicted number of accidents for the period between 2010 and 2070.
nd mandatory ISA are predicted to reduce CO2 emissions by 3.4%nd 5.8% respectively on 70 mph roads. The change in emissionsn non-70 mph speed limit roads is variable and small (Carstent al., 2008). The same proportionate savings can be predicted foruel consumption. Other emissions savings, apart from CO2, wereonsidered to be negligible, as were reductions in noise.
It is assumed that other light vehicles have the same fuel andmissions savings as cars, and it is also assumed that there is nouel or emissions savings for heavy goods vehicles, coaches or buses.he potential of ISA in saving fuel and CO2 emissions was examinedased on the two deployment scenarios. Historical data on fuel con-umption between 2002 and 2006 were used to produce an averagealue, which was then rolled forward to 2010 as fuel consumptionf the base year.
Since the effect of ISA on non-70 mph speed limit roads was rel-tively small, this analysis focused on data applicable to 70 mphoads. As a result, vehicles weighing over 3.5 tonnes were notncluded in the analysis, because these vehicles are limited at a toppeed of 62 mph on roads where a 70 mph speed limit is applicableo cars and light goods vehicles. In addition, it is assumed that lightoods vehicles weighing under 3.5 tonnes have the same fuel andmissions savings as cars. Furthermore, fuel consumption for petrolnd diesel powered vehicles was considered separately, to accom-odate the expected changes in the proportion of future vehicle
eets as well as the differences in the expected improvement inuel efficiency.
Table 6 shows the predicted savings in CO2 emissions anduel consumption over the 60-year appraisal period. The Authorityriven scenario achieved substantially better results than the Mar-et Driven scenario due to its higher proportion of intervening ISAhich delivers more effective speed reductions than advisory ISA.
. Cost benefit analysis of ISA
Following the prediction of ISA in reducing accident occurrence,uel consumption, and CO2 emission, the benefits of ISA were mon-
able 7stimated costs (£) for in-vehicle ISA hardware.
New vehicle
Advisory Voluntary/mandatory
2010 90 200
2020 60 135
2030 60 135
2040 60 135
2050 60 135
2060 60 135
2070 60 135
Cost 1,770,056 205,261 21,068
etised to provide an indication of their economic value. To assesswhere there is a case for proceeding with ISA deployment on thegrounds that the total benefits that accrue over the appraisal periodoutweigh the investment in ISA, it is of course also necessary toestimate the profile of ISA equipment costs over the period and toestimate what additional costs might be incurred by, for example,the need to supply digital road maps incorporating speed limits.
5.1. The costs of ISA
The costs of ISA can be broken up into infrastructure costs(those costs need to create the digital map and perhaps to broad-cast map updates) and the costs of the in-vehicle equipment. Thecosts of compiling a digital road map incorporating speed limitsare comparatively small and the resulting maps would have manyother uses. Keeping the information up-to-date would likely beno more costly than with current manual processes. It is there-fore not considered appropriate to ascribe the cost of compilingthe initial map to ISA. Broadcasting of updates to the maps, sothat the in-vehicle information was up-to-date, would also bequite low-cost, and broadcasting services would be used by othertraffic management applications, so it is not considered appro-priate to ascribe these costs solely to ISA. The major costs of ISAare related to the in-vehicle equipment. It is assumed that, aswith other electronic equipment, ISA costs will fall substantiallyover time, but that beyond 2020 costs will not decline further.Retrofitting of ISA is considerably more expensive that fittingas original equipment on new vehicles, because of the requiredlabour. Table 7 shows the prediction of costs over the appraisalperiod.
5.2. The benefits of ISA
The benefits of ISA in this analysis were attributable to threesources:
Retrofit
Advisory Voluntary/mandatory
247 357250 325282 357330 405380 455433 508503 578
70 F. Lai et al. / Accident Analysis and Prevention 48 (2012) 63– 72
Table 9Price index for crude oil relative to the 2010 price.
Deployment scenario ISA benefits (£m in 2008 prices) Proportion of benefits dueto accident savings (%)
Fuel CO2 Accidents Total
38
nc
•
•
•
aa
•
•
•
•
awv
TF
Market Driven 446 196
Authority Driven 1740 773
Reduction in accident occurrence: this involved monetising theumber of accidents that would be reduced due to ISA. The analysisonsisted of the following steps.
The number of accidents in the base year was derived from theaverage number of accidents recorded in the UK’s national acci-dent database, STATS19, in the years 2002 through 2006, brokendown by severity levels.Predicted reduction in the number of accidents in future years,taken the forecasted traffic growth and the accident rate reduc-tion factor (Department of the Environment Transport and theRegions, 1997; Department for Transport, 2004) into account.Predicted costs of accident (Department for Transport, 2004,2007). Costs at the 2008 price is given in Table 8 and are increasedeach year by the expected increase in Gross Domestic Prod-uct (GDP), offset by appropriate discount rate (Department forTransport, 2004).
Reduction in fuel consumption: this involved monetising themount of fuel consumption that would be reduced due to ISA. Thenalysis consisted of the following steps.
Historical fuel consumption by cars and Light Goods Vehicles(LGVs) between 2002 and 2006 was used as the base year fuelconsumption (Office for National Statistics, 2008).Predicted fuel consumption in future years, taken the trafficgrowth (Department for Transport, 2008) and fuel efficiency fac-tors (Department for Transport, 2003) into account.Predicted cost of fuel taken the forecasted growth in the ResourceCost of Fuel (i.e. excluding fuel duty and Value Added Tax)(Department for Transport, 2003) and fuel price growth scenar-ios (Department for Business Enterprise and Regulatory Reform,2008) into account.Future price offset by appropriate discount rate (Department forTransport, 2003).
Table 9 depicts the four fuel price growth scenarios used in thenalysis. The Resource Cost of Fuel used for the Base Year (2010)as £25.96 for average car and £27.28 for average LGV. These two
alues were used for the Central price scenario.
able 11uel saving by ISA deployment scenario (£m in 2008 prices).
Deployment scenario BERR fuel price scenario
Low Cent
Market Driven 436 446Authority Driven 1703 1740
2,808 33,450 985,122 87,636 97
Reduction in carbon dioxide emissions: this involved monetisingthe amount of CO2 that would be reduced due to ISA. The analysisconsisted of the following components.
• Predicted the volume of CO2, taken forecasted CO2 emission(gram per litre of petrol or diesel) (Department for Transport,2003) and forecasted vehicle fleet proportion (Department forTransport, 2003) into account.
• Obtained the predicted fuel consumption from the previous anal-ysis.
• Predicted the cost by the social value of CO2 (Department forTransport, 2003).
• The social value of CO2 used for the Base Year (2010) was £86.55per tonne of carbon dioxide.
Table 10 depicts the benefits derived from the three sources andthe total benefits under both deployment scenarios. It can be seenthat accident savings far outweigh the values of fuel or CO2 saved.It is worth noting that sensitivity analyses were carried out againstdifferent scenarios of future fuel price (Department for BusinessEnterprise and Regulatory Reform, 2008), as depicted in Table 11.Although different price scenarios led to a considerable disparityamong the predicted amount of savings, the impact of fuel pricevariation does not affect the total benefits of ISA to a great extent,due to the relatively small contribution of fuel savings to the totalbenefits as can be seen from Table 10.
5.3. The benefit-to-cost ratio
Table 12 reports the cumulative benefit-to-cost ratios (BCR)over the appraisal period. The overall BCRs over the whole 60-yearperiod considered are 3.4 for the Market Driven scenario and 7.4for the Authority Driven scenario, based on the BERR Central sce-nario for fuel prices. The costs of deploying ISA tend to be larger
in the earlier years and the benefits tend to come later when morevehicles have ISA. However the payback time is not very long; thebenefits would outweigh the costs by 2025 under both deploymentscenarios.
ral High High High
545 624 2139 2439
F. Lai et al. / Accident Analysis and
Table 12Cumulative benefit-to-cost ratios over time for each implementation scenario.
This analysis demonstrates that ISA is a system with the poten-ial to save a very considerable proportion of accidents. Thatotential is largest on 30 mph roads, where there is considerablepeeding and a large proportion of crashes involve collisions withedestrians. However, on other types of road such as trunk motor-ays and A-roads, ISA also has considerable potential as a safetyeasure. The clear finding from the predictions of the safety impact
f ISA over time is that ISA can have a large effect on future acci-ent number and particularly on the more severe accidents. Overall,dvisory ISA is predicted to be substantially less effective than thentervening (voluntary and mandatory) forms of ISA.
This analysis draws on the data collected in the ISA-UK fieldrials and then applies that data to the prediction of changes in acci-ent numbers through the application of available speed accidentelationships derived from empirical observations. Although theres considerable literature on the impact of speed on accident risk,here is not a single UK based model that is directly applicable toSA. The speed accident relationships selected for the combinationf ISA variants and road characteristics (i.e. Table 2) were consid-red to be the best models available following a series of sensitivityests carried out earlier (Tate and Carsten, 2008).
Based on the modelling results, the impact of ISA is increasedy strategies that lead to higher level of penetration into the vehi-le fleet and also by more rapid introduction of intervening ISA inhe form of the voluntary system. In the Authority Driven scenario,SA is predicted to save 30% of fatal accidents and 25% of seriousccidents over the 60-year appraisal period. The environmentalenefits of ISA are real but become insignificant when comparingith the safety benefits.
It is worth noting that the system variants and deployment sce-arios considered here are not necessarily independent of eachther. Both advisory and voluntary ISA are envisaged in bothcenarios, although in quite different proportions. The Authorityriven scenario envisages that the use of ISA on equipped vehi-les is eventually made compulsory; i.e. upon full penetration ofoluntary ISA, mandatory ISA would be implemented to maximise
he safety benefits. When would be the appropriate time point towitch to mandatory ISA is entirely a modelling assumption, andoes not bear any political implications. It is important to note thatarlier enabling of mandatory usage is technically feasible, but it
Prevention 48 (2012) 63– 72 71
would require faster progress in retrofitting ISA to older vehicles.This would potentially create an attractive market for vehicles notfitted with the mandatory system, which would be undesirable onsafety grounds.
It should also be noted that the cost benefit analysis aimed toconsider the major benefits that ISA could bring, but did not tryto include a comprehensive list of benefits applicable to speedreduction. For example, ISA has major potential to deliver trafficcalming at low cost. In addition journey time saved through exceed-ing the speed limit is considered as a benefit illegally acquired andhence the ‘loss’ of such time through compliance with ISA was notincluded in the cost benefit analysis.
Overall, it can be concluded that implementation of ISA is clearlyjustified from a social cost and benefit perspective. Of the twomodelled scenarios, the Market Driven one is substantially out-performed by the Authority Driven one. This is not altogethersurprising as the benefit side is dominated by the accident savingswhere voluntary ISA is far more effective than advisory ISA.
Acknowledgements
This work was carried out within the ‘Speed Limit Adherenceand its Effect on Road Safety and Climate Change’ project funded bythe Commission for Integrated Transport (CfIT) and the Motorists’Forum (MF). The opinions, findings, and conclusions expressed inthis paper are those of the authors alone and do not necessarilyreflect those of CfIT and MF.
References
Aarts, L., van Schagen, I., 2006. Driving speed and the risk of road crashes: a review.Accident Analysis and Prevention 38 (2), 215–224.
Department for Business Enterprise and Regulatory Reform, 2008. Communicationon BERR Fossil Fuel Price Assumptions. Department for Business Enterprise andRegulatory Reform., London.
Carslaw, D.C., Goodman, P.S., Lai, F.C.H., Carsten, O.M.J., 2010. Comprehensive anal-ysis of the carbon impacts of vehicle intelligent speed control. AtmosphericEnvironment 44 (23), 2674–2680.
Carsten, O., Lai, F., Chorlton, K., Goodman, P., Carslaw, D., Hess, H., 2008. Speed LimitAdherence and its Effect on Road Safety and Climate Change. Commission forIntegrated Transport, London.
Department for Transport, 2003. Transport Analysis Guidance.http://www.dft.gov.uk/webtag/ (initially published in 2003).
Department for Transport, 2004. Design Manual for Roads and Bridges, vol. 13. Eco-nomic Assessment of Road Schemes, Section 1: The COBA Manual. Departmentfor Transport, London.
Department for Transport, 2007. Road Casualties Great Britain 2006. Departmentfor Transport, London.
Department for Transport, 2008. Road Transport Forecasts 2008. Department forTransport, London.
Department of the Environment, Transport and the Regions, 1997. National RoadTraffic Forecasts. Department of the Environment, Transport and the Regions,London.
Duynstee, L., Katteler, H., Martens, G., 2001. Intelligent speed adaptation: selectedresults of the Dutch practical trial. In: Proceedings of the 8th ITS World Congress,30 September to 4 October, Sydney.
Ehrlich, J., Saad, F., Lassarre, S., Romon, S., 2006. Assessment of “LAVIA” speed adap-tation systems: experimental design and initial results on system use and speedbehaviour. In: Proceedings of 13th ITS World Congress , London, 8–12 October.
Elvik, R., Christensen, P., Amundsen, A., 2004. Speed and road accidents: an evalua-tion of the Power Model. TOI Research Report 740/2004. Institute of TransportEconomics, Oslo.
Hjälmdahl, M., Várhelyi, A., 2004. Speed regulation by in-car active accelerator pedal:effects on driver behaviour. Transportation Research Part F: Traffic Psychologyand Behaviour 7 (2), 77–94.
Hastie, T.J., Tibshirani, R., 1990. Generalized Additive Models. Chapman and Hall,London.
Kloeden, C.N., Ponte, G., McLean, A.J., 2001. Travelling Speed and the Risk of CrashInvolvement on Rural Roads. CR 204. Road Accident Research Unit, Universityof Adelaide, Australia.
loeden, C.N., McLean, A.J., Glonek, G., 2002. Reanalysis of Travelling Speed and theRisk of Crash Involvement in Adelaide South Australia. CR 207. Road AccidentResearch Unit, University of Adelaide, Australia.
ai, F., Chorlton, K., Carsten, O., 2008. Overall Field Trial Results. Report of the Intel-ligent Speed Adaptation Project. Institute for Transport Studies, University of
Leeds.
ahrmann, H., Madsen, J.R., Boroch, T., 2001. Intelligent speedadaptation—development of a GPS based ISA systems and field trial ofthe systems with 24 test drivers. In: Proceedings of the 8th ITS World Congress,30 September to 4 October, Sydney.
Prevention 48 (2012) 63– 72
Nilsson, G., 1982. The Effect of Speed Limits on Traffic Accidents in Sweden.Report 68 National Swedish Road Traffic Research Institute (VTI), Linköping,Sweden.
Office for National Statistics, 2008. Energy uses by economic sector, source and fuel
Tate, F.N., Carsten, O.M.J., 2008. Implementation Scenarios. Report of the IntelligentSpeed Adaptation Project. Institute for Transport Studies, University of Leeds.
Taylor, M.C., Lynam, D.A., Baruya, A., 2000. The Effect of Drivers’ Speed on the Fre-quency of Road Accidents. TRL Report 421. TRL Limited, Crowthorne, UK.
