[10-01-08] Vacature student assistent voor AIDA STOK pilot

Vacancy for

Two student assistants at the research centre

Applications of Integrated Driver Assistance (AIDA).

In 2008 the AIDA research centre will start a project on modelling the impact of incentives on route choice and traffic safety using a route choice simulator experiment. For this project we are looking for two student assistants, each for 0,5 fte during 2008.

The first student assistant will work on the adaptation of the route choice simulator and implement the incentive structure and safest route algorithm. We require good programming skills (in particular Java).

The second student assistant will work on the experimental design, conduct the experiment and estimate route choice models on the basis of experimental results. We require a background in traffic and transport and an interest in traffic and transport models.

You like to work in a multidisciplinary work environment. You are able to communicate both orally and written in English.

We offer the standard conditions for student assistants at the University of Twente.

More information:

·

enclosed project description

·

www.aida.utwente.nl

·

prof. Bart van Arem, 053 4893046; b.vanarem@utwente.nl

Modelling the impact of incentives on route choice and traffic safety using a route choice simulator experiment

Thijs Muizelaar

Bart van Arem

Introduction

In the recent years a number of studies and experiments have been done towards the varying of the insurance premium. Around the world, a number of insurance companies have introduced such a policy, in which the amount of kilometres driven each year is the basis for the insurance premium. In the future, these car owners will pay their insurance by the distance they travelled. However, for the professional driver, no such policy is available. Especially for companies such as couriers, it is very expensive to get car insurance, because of the high risks for the insurance company. Varying the insurance premium per kilometre might not work for such companies, but the current technology offers more options to make the insurance premium variable.

An insurance company aims at receiving enough premiums from its insured vehicles, while having as little claims as possible. Reducing the number of claims is thus profitable for the company, and should also be profitable for the customer. If the insurance company can influence the driving behaviour and choices of the customer, the premium could be reduced according to the changed behaviour. To control the behaviour of the customer, it is necessary to measure his actual driving behaviour, and his other choices, such as routes and time of travel. If the customer chooses to travel during periods and over routes which have a reduced risk of accidents, the insurance premium could be lower. The same applies for someone who shows a very safe driving behaviour. In other words, the customer pays the insurance premium that fits his driving behaviour.

In this study we are interested in the effects of providing advice for the safest route to professional drivers, while also providing feedback to the driver about his driving behaviour. As a trigger to follow the advice, the insurance premium is made variable. The company, at which the driver works, could then receive a reward or incentive if his behaviour leads to a lower premium. When a driver acts according to the suggestions, the insurance premium will be lower (STOK, 2007, TNO, 2003).

Literature

Varying the insurance premium is not yet a common way to insure a vehicle. Only a few companies offer such an insurance policy (Norwich Union in the UK, Progressive Insurance in the USA, Holland Insurance in South-Africa and a few others (VTPI, 2007 & STOK, 2007)). Most of these insurance policies are based on a premium per kilometre drive and are also called Pay-As-You-Drive (PAYD) insurance policies. The premium of the insurance is, apart from certain risk characteristics such as age and region, based on the amount of kilometres driven.

A number of ideas are the basis for this type of car insurance (VTPI, 2007, Litman, 2006a, 2006b, Guensler & Ogle, 2001). First of all, it is expected (based on price elasticity) that the amount of kilometres drive is reduced, because drivers have to pay for using a vehicle, while the insurance premium currently is paid up front. Trough this reduction most car owners reduce the cost of using their vehicle. The reduction also brings forward a reduction in external effects, such as accidents, congestion, emissions and use of car fuel. Another idea behind PAYD is that it leads to a fairer distribution of car ownership and car use, such that lower incomes also can afford to own a car.

Besides the advantages of PAYD, a number of critical annotations can be made (VTPI, 2007, Litman, 2006a, 2006b, Guensler & Ogle, 2001). Most PAYD policies only use the amount of kilometres driven. Apart from this variable many others exist which also have a great influence on the chance of an accident, such as driving behaviour. These variables should also be taken into account. The possibility that the reduction in kilometres driven are the kilometres with the lowest chance of an accident also exists. The reduction then does not lead to a reduction in the number of accidents (or claims). The privacy of a driver is also worth some attention, especially when the driving behaviour is taken into account. As long as only the amount of kilometres driven is used, no problem exists with privacy, because this information is currently already registered with each annual inspection (APK).

From the existing literature a number of notable items arise. First, the focus of all the studies is aimed at the non-professional driver. Freight transport is not at all taken into account as a possible party of interest for a PAYD insurance policy. The results so far cannot easily be transferred to a professional environment, because the assessment of costs is different. After all, a non-professional driver has to pay for the insurance himself, while a professional driver has the owner of the vehicle (other than the driver) paying for the insurance. Influencing the driving behaviour of a professional driver using a PAYD insurance thus is not obvious.

Second, most PAYD insurance policies are based on the amount of kilometres driven. The policy does not take into account in which way these kilometres are driven, via which route, at what time of day or in what area. Mostly the existing criteria for riskgroups are used, such as age and residence. If a GPS device is used, it would be possible to use these other variables as a basis for the PAYD insurance premium.

Third, all existing PAYD policies only aim at varying the insurance premium, actually targeting a reduction of the kilometres driven. This causes the chance of an accident happing to be reduced, causing fewer claims for the insurance company. No effort is done to influence the driving behaviour directly. Only the insurance premium varies. A driver only knows at the end of the month what his premium will be. With the aid of a personal navigation system, it would be possible to directly influence the driver, for example by advising the cheapest or safest route.

The existing literature shows some information on the effects of PAYD and the height of the insurance premium based on estimations (VTPI, 2007, Litman, 2006a). For the USA, the premium would on average be 6 cents per mile. This would mean a reduction of driven miles of around 10%, based on the price elasticity. For a number of regions in the USA a calculation was made what the effect would be with a premium of 2 cent per mile. That would lead to 4% less driven miles. In most cases the premium is calculated using the current premium divided by the driven miles in a year. Someone with currently a low premium would get a low premium per mile, and based on price elasticity, would show only a small reduction in miles driven. A higher premium per mile would lead to higher reductions. On the basis of accident statistics it was estimated that a reduction of 10% in miles would lead to a reduction of 17% in accidents, because a vehicle itself has less change to cause an accident, but also gets less often involved in an accident caused by other vehicles.

To speedup the introduction of variable premiums a market research was done in Minnesota, USA (Buckey et al., 2007). The research showed that only a small group is in itself interested in a variable premium, although they were interested in only paying for using a vehicle. The most important barrier that are mentioned are the uncertainty of the costs and the privacy. It also showed that most drivers have no idea of the price per kilometre of using a vehicle.

When looking at the utility of reducing the amount of miles driven, it shows that for each mile a utility can be found of around 16 cent, of which half is private and the other is public (Greenberg, 2007). This could be used to speedup the investments necessary to introduce PAYD insurance policies.

The relationship between route choice and safety is not often found in the available literature. Only a Greek study (Yannis et al., 2005) shows that non-professional drivers make a choice for a safer route on the basis of traveltime related parameters. Costs are not important for choosing a safer route. This would mean that varying the premium using an advice for the safest route would not be very useful. Because the pilot aims at the professional driver, for whom costs play a very different role, it does deserve attention. Other important characteristics that play a role in choosing a safe route were gender, income and driving experience.

Conceptual model

The conceptual model can be found in Figure 1. On the right side of the model, the available human actors are depicted. These actors are:

·

the driver;

·

the planner;

·

the haulier or entrepreneur;

·

the insurance company.

The relation between these actors is made dynamic with a varying insurance premium. Because of the professional setting, the premium is not directly paid by the driver, but more likely by the owner of the company, or in other words the haulier or entrepreneur. This actor has to deal with both the driver and the planner, in order for the varying insurance premium to work. These two actors namely have a large influence on the route choice. This relation is depicted with the variable reward or incentive. For the insurance company and the haulier, it means an “improvement” of the service provided. This service can be individualized, allows for a more sustainable approach and has some operational and financial advantages.

The other elements in Figure 1 depict the more physical elements in the system. Halfway are the adjustable parameters, or in other words the insurance policy that is chosen. This policy can be adapted by the insurance company, but also by the haulier and the planner, because they have their influence on the running of the company. The policy determines the type of route guidance that is given, but also the incentives for the planner and driver.

Figure 1: conceptual model of the system

Figure 1: Conceptual model of the system

The actual route guidance that is given to the driver is based on these adjustable parameters. The driver than acts upon the advice and chooses routes and shows his driving behaviour. Both the route choice and driving behaviour are measured. The driver receives direct feedback on both, and has knowledge of the effects on his reward. The measurements are also saved in a database which is available for all other actors, which allows them to act accordingly (adjust the parameters).

Work packages

The study consists of the following work packages.

1.

Definition

2.

Specification

3.

Pilot preparation

4.

Pilot execution

5.

Pilot analysis

6.

Route choice study

7.

Traffic simulation study

8.

Stakeholder analysis

9.

Final report and symposium

10.

Project management

In work package 1 and 2 the incentive structure will be developed in more detail. Work package 3-5 pertain to the actual pilot study and will not be described here. Work package 6 will be described in some more detail, as it will be the main task of the student assistants. Work package 7 will be using results of work package 6. Work package 8 will develop models for stakeholders preferences. Work package 9-10 speak for themselves.

Route choice study

The goal of this work package is to perform a route choice study to investigate the effects of the incentive structure (and variations) and the safe route guidance on the route choice of professional drivers.

Little is known about the effects of varying types of route guidance on the actual route choice of drivers, let alone professional drivers. Also the influence of an incentive on the route choice is unknown. In this work package the effects will be investigated and put into a route choice model for further use.

A Greek study investigated the choice of drivers to reduce the chance of an accident (Yannis, et al., 2005). Using a questionnaire they found out that variables related to the total traveltime, extra traveltime and costs are of importance on the choice for a route with less chance of accidents. Especially the extra traveltime is a very important parameter for choosing a safer route. Characteristics other than those of the route itself, which are important, are the gender and driving experience.

Acquiring the necessary data regarding the route choice in the pilot experiment can be done using a GPS device which saves the location of the vehicle at the proper frequency. By matching the measured location with a map it is possible to define the routes of that vehicle. GPS has been used before to estimate route choice models (Rich, et al., 2007, Frejinger & Bierlaire, 2006, Bierlaire & Frejinger, 2007). For a more controllable experiment a virtual environment can be used, such as a route choice simulator (Raadsen & Muizelaar, 2005, Bonsall et al., 1997). In a route choice simulator are testpersons asked to make a journey trough a roadnetwork, where they can indicate their choices using a computer. They can receive various types of traffic information, as well as route guidance. The traffic situations they encounter can be different each time and can be presented in various ways to the testperson. The data from a route choice simulator is also a possible source for estimating route choice models.

With the aid of the route choice data it is possible to develop a route choice model, based on the Random Utility Maximizing theory (Rich, et al., 2007, Frejinger & Bierlaire, 2006, Bierlaire & Frejinger, 2007). It is possible to use different incentive structures to see what their effects is on the route choice behaviour. A model also enables extrapolation to untested structures, and application in a traffic simulation model.

Tasks and activities

This workpackage aims at modelling the route choice behaviour. To be able to do so, it is necessary to have the right data available. First part of the workpackage thus is the experimental design, for both the pilot and the route choice simulator study. Following the design, the route choice simulator should be adapted. The next task is the actual pilot and experiments in the simulator. Using the data a route choice model can be estimated. Finally the results should be put into a modeling environment for further simulations, and a report should be made

Schedule

The following preliminary schedule applies.

Period

Task

November 2007 – December 2007

Develop experimental design for the pilot study and the route choice simulator study.

December 2007 – March 2008

Adapt the route choice simulator and implement the incentive structure and safest route algorithm.

April 2008 – June 2008

Perform the route choice simulator experiment.

April 2008 – April 2009

Perform the pilot experiment.

June 2008 – August 2008

Estimate route choice models based on simulator data.

August 2008 – October 2008

Implement route choice models in simulation environment (ITS Modeler).

October 2008 – December 2008

Write report on the route choice models.

April 2009 – July 2009

Estimate route choice models based on pilot data.

July 2009 – September 2009

Write report on route choice models.

Organization

·

Prof Dr Bart van Arem (general supervision)

·

Thijs Muizelaar

·

Dr Jing Bie (also responsible for step 5 in this study)

·

Student assistant implementation route choice simulator, reporting

·

Student assistant experimental design, experiments, analysis and estimating route choice models, reporting.

We expect that both student assistants will work half-time (average) during 2008. At the start of 2008 the implementation may take more than half time. Medium 2008, the experiments may take more than half time. The daily supervision will be done by Dr Jing Bie.

Literature

Victoria Transport Policy Institute, 2007, Pay-As-You-Drive Vehicle Insurance, http://www.vtpi.org/tdm/tdm79.htm, Visited on 1 October 2007

Todd Litman, 2006, Mobility Management Traffic Safety Impacts, In proceedings TRB Annual Meeting 2006, Paper 06-1475, Washington, USA

Todd Litman, 2006, Pay-As-You-Drive vehicle Insurance, In proceedings TRB Annual Meeting 2006, Paper 06-1796, Washington, USA

Kenneth R. Buckey, et al., 2007, Minnosota Pay-As-You-Drive Market Research, In proceedings TRB Annual Meeting 2007, Paper 07-1687, Washington, USA

Allen Greenberg, 2007, Designing Pay-Per-Mile Auto Insurance Regulatory Incentives Using the NHTSA Light Truck CAFÉ Rule as a Model, In proceedings TRB Annual Meeting 2007, Paper 07-3457, Washington, USA

Randall Guensler & Jennifer Ogle, 2001, Commuter choice and value pricing insurance incentive program, Project proposal for the Value pricing Pilot Program, Georgia Institute of Technology, Atlanta, USA

TNO Inro, 2003, ‘Pay as you drive’ in Nederland, Documentatiebladnummer 2003-04, Delft, The Netherlands

STOK, 2007, http://www.stok-nederland.nl, Visited on 3 October 2007

De Ridder, S.N., & Hoedemaeker, M. (2003). LDWA 1: behavioural effects of Lane Departure Warning Assistance (LDWA) Systems in a Field Operational Test (report TM-03-C034). Soesterberg, The Netherlands: TNO Human Factors.

Mazureck, U. & D. de Wit (2006). Belonitor de kracht van belonen, Delft, Netherlands, ISBN 90 3693 632 2, in Dutch

Mazureck, U. (2006). Rewarding Safe Driving Behavior: Influencing Following Distance and Speed. Transportation Research Board 85th Annual Meeting, Washington D.C., January 2006

Vonk, T., Van Rooijen, T., Hogema, J., & Feenstra, P. (2007). Do navigation systems improve traffic safety ? (TNO report 06.34.15/N121/TVo/LK). Delft, The Netherlands: TNO.

G. Yannis, et al., 2005, Model driver choices towards accident risk reduction, Safety Science no. 43, pp. 173-186

J. Rich, et al., 2007, Route choice model for Copenhagen, In proceedings Tristan 2007, Phuket, Thailand

E. Frejinger & M. Bierlaire, 2006, Capturing correlation with subnetworks in route choice models, Transportation Research Part B, Vol. 41, No. 3, pp. 363-378

M. Bierlaire & E. Frejinger, 2007, Route choice modeling with network-free data, Transportation Research Part C, in-press

M.P.H. Raadsen & T.J. Muizelaar, 2005 Ontwikkeling van een dynamische routekeuze simulator, Colloquium Vervoersplanologisch Speurwerk 2005, Antwerpen

P. Bonsall et al., 1997, Validating the results of a route choice simulation, Transportation Research Part C, Vol. 5, No. 6, pp. 371-387