Safety, mobility and comfort assessment methodologies of intelligent transport systems for vulnerable road users

Summarising the extensive literature review of [14], the following methods can be distinguished for an ex-ante and ex-post evaluation of active safety systems:

The methodologies reviewed above have hardly been applied to vulnerable road users. Methods based on ex-post analysis of accident statistics are not relevant for the case of ITS for vulnerable road user, because little or no statistical data on such applications are available. The lack of statistical data also precludes the use of models with “predictive” ability, using data about the past. This is due to the ITS being a “trend break” rather than a measure for which the impacts can be predicted. Thus, ex-ante assessment methods seem more appropriate. A comprehensive approach that covers all possible safety effects is the safety mechanisms method, and therefore this was used in the VRUITS project.

The starting point for the development of the safety impact assessment method which was used in VRUITS project was the safety impact assessment framework presented by [15]. The framework of [15] is based on the theoretical background presented by [19] according to which the traffic safety consists of three dimensions, which are (1) exposure, (2) risk of a collision to take place during a trip and (3) consequences (= risk of a collision to result in injuries or death) ([19], as illustrated in Fig. 2). The volume of the rectangular box is the expected number of injured or fatalities. Thus the number of injuries or fatalities in road accidents depends on the three dimensions – exposure, accident rate and injury severity.

The framework of [15] emphasises the systematic nature of transport: when one element of the system is affected, the consequences may appear in several elements and levels of the system. Therefore, the implemented measures influence safety by affecting one or several of the factors contributing to any of these three dimensions of safety. The use of this approach ensures that the safety impact assessment method will cover all dimensions of road safety, also exposure or the amount of travelling, which is frequently overlooked in the safety assessment studies [15].

In addition to the three dimensions of road safety (as indicated in Fig. 2) the framework for the safety impact assessment of ITS should also cover the effects due to behavioural adaptation in addition to engineering effects, and be compatible with other aspects of state-of-the-art road safety theories [15]. In order to be sure that all possible impacts (both positive and negative impacts on road safety; direct, indirect and unintended effects of systems) will be covered, the analyses proposed by [15] utilises a set of nine mechanisms via which ITS can effect road user behaviour and thereby road safety. These nine mechanisms cover the three aspects of road safety in a systematic manner and are based on a ten-point list compiled by [3]. For the purposes of the VRUITS project, these nine mechanisms were updated to cover vulnerable road users, i.e. pedestrians, cyclists, moped riders and motorcyclists. The updated mechanisms, which are now more focused on the changes in behaviour of vulnerable road users and the situations they face in traffic [29], are presented below:

As indicated by [15] many of these mechanisms are closely linked to one another, and could be combined. Examples of these are mechanisms of direct driving behaviour modification (1–2), indirect driving behaviour modification (3–5), and travel pattern modification (6–8). The mechanisms have not, however, been combined since the purpose of the framework is to illustrate the types of different possible effects (both positive and negative) of IT systems on safety.

In the VRUITS project the accident data was handled similarly as in the INTERSAFE2 project [30] and in the eIMPACT project [30]. The accident data in the eIMPACT project covered the EU-25 countries from 2005 [31] and the INTERSAFE2 project updated the accident data to include the new member countries Romania and Bulgaria and extended the database to the EU-27. In the VRUITS project the accident data was further extended to cover the EU-28. The CARE database [2] was chosen for the analysis due to its European coverage. The total number of fatalities and injuries used in the calculations are presented in Table 2. The figures for 2020 and 2030 were calculated based on accident trends including separate estimates for accidents related to pedestrians, cyclists, moped riders, motorcyclists and cars [6].

The total number of fatalities used in the impact assessment calculations for the EU-28 was taken from the Statistical pocketbook [4]. The statistical pocketbook does not include any information on the number of injuries (only on the number of injury accidents) and thus the total number of injuries was taken from CARE database. Based on our analysis, the annual number of fatalities reported in CARE in 2012 (25,738) matched well with the number reported in the Statistical pocketbook (25,776), when taking into consideration only the countries for which 2012 data were available. Therefore we made the assumption that the number of injuries reported in CARE matched well with the numbers in the statistical pocketbook and no further correction was necessary.

More detailed information on fatalities and injuries for the EU-28 were gathered from the statistics of CARE database for the year 2012. No accident data for 2012 were available for Belgium, Bulgaria, Estonia, Lithuania, Malta, Slovakia and Sweden and thus the latest available data in CARE database was used for those countries instead (2011 for Belgium, 2010 for Malta, Slovakia and Sweden and 2009 for Bulgaria and Estonia). The CARE database included no information on road accidents for Lithuania. Thus the total numbers of fatalities and injuries in 2012 were taken from the Lithuanian national statistics. The resulting figure for fatalities in the EU-28 was 28,126.

The data from the CARE database were used to classify the fatalities and injuries according to the following background variables: collision type, road type, weather conditions, lighting conditions, location of the accident and age. Background variables, sometimes called “situational variables”, are necessary because they provide insight into whether an ITS is appropriate for certain circumstances. If not, the effectiveness of the ITS in that situation is negligible or nil. Compared to the previous projects (e.g., eIMPACT and INTERSAFE2), it was not possible to determine if the vulnerable road user had been in an accident with a heavy or light vehicle (except for pedestrian accidents). This is because CARE does not allow for the collision partner to be identified; the data are classified using one vehicle type e.g., it is possible to tell that a cyclist had an accident with multiple vehicles but it is not possible to distinguish the characteristics of the other vehicles. The distribution of fatalities and injuries according to the different background variables was exploited when calculating the safety effects of different ITS (see Section 5).

There is variability in the quality of the accident data entered into European-wide accident databases by country with some being highly detailed and accurate whereas others have many cases of ‘unknowns’. To generate background variable data, the following approach was used, similar to eIMPACT. This approach groups countries with similar safety characteristics together. The accident data for the EU-25 in eIMPACT project [30] was divided into three clusters. The clusters were formed based on the prevalent safety situation in each country and therefore the countries with similar road safety situations were included in the same cluster. For the countries where no detailed information was available on the background variables, or when the values were not considered reliable, the average values from the cluster to which the country belonged were used. In VRUITS, these clusters were updated by using the latest road safety and vulnerable road user safety-related statistics.

Tables 3 and 4 present the distribution of vulnerable road user fatalities and injuries by collision type (Table 3) and the distribution of vulnerable road user fatalities and injuries by other variables (Table 4). The total number of fatalities and the total numbers of fatalities of pedestrians, cyclists, moped riders and motorcyclists were taken from the statistical pocketbook from which the share of single vehicle accidents were separated based on the information reporting in CARE database. For this work, only the information from countries that had provided the specific information was taken into consideration. Regarding injuries, the classification of injuries into different collision types was based exclusively on the data reported to the CARE database for the countries that had provided the specific information. The adverse weather conditions presented in Table 4 included fog, mist, rain, snow, sleet and hail. Twilight was included in the lighting condition Daylight. Compared to previous projects, age group was included in the analysis as an additional variable.

In the VRUITS project, the effects of exposure (as part of Mechanism 6) focused on changes in exposure as measured in kilometres of travel for different vulnerable road user groups. The estimated effects on vulnerable road user exposure were transferred to safety effects of exposure (the same values were used for fatalities and injuries) based on the values found in earlier studies. These studies found a “safety in numbers” effect, where the accident risk per kilometre for vulnerable road user modes decreases as vulnerable road user travel increases – a likely explanation is that other road users learn to expect vulnerable road users on the road when they become less rare. This means that the number of fatalities and injuries increases slower than the number of vulnerable road user kilometres. The studies modelled this with an exponential model: if the exposure increased by a factor of x, then the number of fatalities and injuries increased by a factor x^y, for some exponent y less than 1. The exponent y was 0.38 for the pedestrians (based on [12]), 0.4 for cyclists and moped riders (based on [11]) and 0.7 for motorcyclists (based on [17] and previous impact assessment studies). The safety effects of modal change (Mechanism 7) were calculated by using the same formulas as for the change (increase or decrease) in the exposure of vulnerable road users (Mechanism 6). The mechanism-based approach in our safety assessment allowed us to conclude that no double counting of the effects occurred. The effects of the modal shift were only calculated for vulnerable road users; the effects of the modal change of cars, trucks and public transport were ignored. A small change in the modal share of vulnerable road users corresponds to a significantly smaller change in the modal change of cars, trucks and/or public transport, thus the safety effect of the modal shift to or from cars, trucks and public transport would be negligible. Moreover, the risk per km for cars, trucks and public transport is much smaller than the risk per km for vulnerable road user [5]. Thus, calculating the effect of modal change for cars, trucks and/or public transport results in a smaller change in travelled km that would be multiplied by a smaller risk factor per km.

The method to assess the safety impacts of ITS on vulnerable road users is based on the method introduced by [15], which was developed for the assessment of safety impacts of ITS for cars. This method is aimed at ITS – as was VRUITS − and is comprehensive in its approach, covering all three dimensions of road safety − exposure, crash risk and consequence, the effects due to behavioural adaptation in addition to the engineering effect (effect on target accident contributory factors) and is compatible with the other aspects of state-of-the-art road safety theories. Some parts of the method were enhanced and adjusted to take also into consideration the vulnerable road users. The main modifications of the method for the purposes of the VRUITS project were related to: i) nine mechanisms which were updated to cover vulnerable road users i.e. pedestrians, cyclists, moped riders and motorcyclists, ii) the safety impact assessment tool which was updated to include more detailed information on accidents involving vulnerable road users, iii) accident types and circumstances such as age, road layout and lighting which were considered in more detail when relevant for vulnerable road user and when feasible, iv) the calculation of safety effects of exposure changes and v) the expert assessment which was used to enhance the value of estimates for the nine mechanisms (see section 5).

While mobility levels, levels of service and general mobility behaviour have been very active research fields, resulting in an increasing number of data sets available on both the national and international levels, the concept of comfort for pedestrians, cyclists and Power-Two-Wheelers is often not dealt with explicitly and in detail. Nevertheless, there are a number of studies from the fields of spatial planning, architecture and civil engineering that specifically address this topic. No clear and definitive mobility and comfort assessment methodologies for vulnerable road users are available.

There are a number of different definitions of mobility for different road users. VRUITS used the definition of mobility based on the work of [24], that links well to vulnerable road users and different functionalities of ITS: “vulnerable road user mobility is any form of outside (of the house) movement based on the identified soft transport modes: walking, cycling or motorcycling. These forms of movement are defined by trips from a starting point to a destination (where the destination can also be a public transport stop or station) in order to conduct an out-of-house activity.”

Based on the literature search and review of current studies in the mobility and comfort research field, there is a severe lack of both theoretical and empirical discussion of the comfort topic. Below, the aspects that need to be taken into account when assessing comfort are explored, followed by the explanation of the methodology to address these aspects.

A definition that is relevant especially in view of the implications of ITS solutions for vulnerable road users, comes from Slater [27]. This definition of comfort is: “„(...) a pleasant state of physiological, psychological, and physical harmony between a human being and the environment” ([27], p. 4). This definition identifies three dimensions – physiological, psychological and physical - relevant for the assessment of comfort according to Slater. Each of these three dimensions must be defined and assessed separately, in order to take the complexity of this issue into account.

A definition of the three dimensions, specifically addressing pedestrian comfort, can be found in [25], where the comfort needs of pedestrians in high density and high complexity urban scenarios are discussed. Sarkar pleads for a two-level approach for the assessment, on a micro and a macro level ([25]: pp. 6): the macro level encompasses the general circumstances and the infrastructural context including relevant standards and criteria, referred to as “service levels”. The micro level focusses on the actual quality of the task and on factors that directly influence the individual perceptions referred to as “Quality levels”. This broadens the definition by Slater, as the micro level specifically addresses individual circumstances under which a certain mode is used or has to be used and allows for a more specific assessment of the actual quality of certain situations. This indicates that comfort assessment varies by the transport mode used, as there are specific differences concerning the physical, psychological and physiological work load. To assess work load appropriately, infrastructural, societal and individual circumstances need to be taken into account for the respective mode. While there is also a reference to the accommodation of pedestrian needs and the psychological level of comfort, no specific tools to assess this dimension other than walking speeds and the ability to engage in social interactions are specifically discussed.

A study focussing on the qualitative aspect of (pedestrian) comfort, while defining comfort similarly, includes the emotional component more explicitly:

“(…) comfort for pedestrians is a positive emotional reaction to external surroundings (the walking environment) in different situations, including physiological, physical, social and psychological reactions.” ([21]: p. 2). The emotional component of comfort is considered as: “(…) short-lived emotional reactions rather than cognitive reflections (…)” ([21], p. 2). This approach implies a stronger focus on individual assessments of comfort that is influenced by both external and internal factors. The actual questions that [21] included in their study focus on “thermal comfort, visual comfort, acoustic comfort, tactile comfort, smells, air pollution and allergens, the ease to move and the feeling of security” ([21], p. 2). They also introduce the concepts of efficiency and perceived safety which opens the discussion not only to infrastructural aspects and indicators such as walking speed, but to characteristics that help to specify certain needs in relation to individual preferences and (in)abilities.

The research project PROMPT, focussing on pedestrian comfort needs, shows the importance of considering individual characteristics such as age, gender and state of health, especially in the connection with mobility impairments, as essential factors for assessing comfort.

The aspects of comfort that the evaluation methodology identified are:

The relationship between these aspects are:

Thus, a method must address the quality and service levels as defined above in order to assess comfort.

A model provided by [15] gives insight into frameworks on how to generally assess mobility, focussing specifically on motorised transport. Units of measure, modes, performance indicators, consumer benefits as well as land use, and improvement strategies are also viable for the assessment of vulnerable road user mobility. Examples include person miles travelled, the number of person trips, and travel convenience.

The methodology has to consider all the relevant aspects of comfort. The comfort level of different vulnerable road user groups should be assessed and the specifics of each vulnerable road user group should be taken into account appropriately. A combination of qualitative and quantitative instruments was envisaged.

A combination of approaches for assessing comfort has been developed by [18, 21] for pedestrians.

[21, 22] also applied survey methods such as focus group interviews and in-depth interviews to cover these aspects. In addition, they applied quantitative checklist methods to rate available infrastructure (i.e., pavement conditions, continuity of sidewalks/cycle paths, seatings, etc.) on different scales. This combined approach that integrates both survey methods for individual road user ratings and observations for infrastructure assessments and road user behaviour, provides insight into the quality and service levels that determine the comfort of different vulnerable road user groups.

When trying to connect general concepts of comfort, such as walkability, cyclability, etc. to ITS solutions and their potential impacts on the travel comfort of vulnerable road users, not only systems directly aimed at improving comfort are relevant but also those that are focussed on improving safety. The studies above show that objective as well as perceived safety play an integral role in the individual comfort perceptions. Systems that allow vulnerable road users to identify potentially critical scenarios in traffic and to avoid them, by being warned, routed or re-routed based on current traffic situations and/or general conditions, are directly related to the comfort concept.

The challenges that need to be addressed to develop the mobility and comfort impact assessment methodology are similar to those of the safety impact assessment with respect to the focus on vulnerable road users. These are:

The TeleFOT framework [10] for assessing mobility and comfort identified characteristics in transport that influence mobility (amount of travel and travel patterns) and comfort (journey quality). It provides a structure for assessment. The TeleFOT framework is shown in Fig. 3. The aspects of mobility and comfort are embodied in this framework. Usefully, it identifies indicators that can be quantified, as shown in the boxes at the left hand side of Fig. 3. The TeleFOT approach provides the starting points for the assessment of each ITS: function, design, use case, and types of impact.

The nine mechanisms in the eIMPACT methodology are used to guide the assessment in covering all aspects of the potential type of impact. The aspects from the general description of mobility and comfort are investigated by the mechanisms as described below:

For mobility:

For comfort, workload, stress, uncertainty and safety perception are addressed by mechanisms 1 and 2.

Mechanisms 3, 4 (relevant for comfort) and 5 (relevant for comfort and mobility) are indirect and can be positive or negative.

Mechanism 9 is not relevant.

The method to assess the mobility and comfort impacts of ITS on vulnerable road users was developed from the method introduced by Kulmala (2010), initially developed for the assessment of safety impacts of ITS for cars. The method for mobility and comfort impact assessment followed the steps defined by Kulmala (2010), but when assessing mobility and comfort numbers 1–5 (comfort) and 6–8 (mobility) were of interest. The method was enhanced and adjusted to take into consideration the mobility and comfort of vulnerable road users covering: mobility and exposure of vulnerable road users measured in: trip length, duration and frequency; and comfort of vulnerable road users measured in: how the users perceive their travel, i.e., change their opinion regarding the comfort of the travel undertaken, and perceived safety in relation to traffic.

The safety assessment conducted in the VRUITS project was the first time that the safety impact assessment framework of [15] was applied to calculate the safety effects for vulnerable road users. This project primarily assessed the effects of new systems which have not been yet used in real traffic (systems were not in production or on the market). Effort was invested in creation of detailed descriptions of the functionalities of each system and several experts were involved in finalising the description and specifying the functioning of each system. Next, literature was reviewed to develop valid and reliable estimates of the effects. Usually it was difficult to find any direct evidence about the effectiveness of the systems and therefore we needed to combine several types of evidence from different sources to understand the user context and road user’s potential reactions to systems. To increase the reliability of the estimates, several experts contributed to the work in parallel and the estimates were cross-checked.

Numerical estimates were produced by using the European Risk Calculation tool ERiC whose structure and content was specifically modified for the assessment of safety effects for vulnerable road users. The main modifications concerned the collection and modification of European-wide accident data in order to calculate the effects for the EU-28. The assessment exploited the statistics from the CARE database, but the content was not detailed enough for our purposes without additional modifications. The variability in the quality of the accident data entered into CARE by country was tackled by grouping the countries into three clusters which were formed based on the prevalent safety situation in each country (countries with a similar safety situation were included in the same cluster).

The method is a systematic approach to cover all effects of the systems; including not only the expected positive effects of the systems but also the indirect and negative effects. Compared to previous applications of the safety assessment method of ITS [15], mechanism 9 (modification of accident consequences) was not relevant for the systems for vulnerable road users in our analysis. Mechanism 9 only applies to systems that affect the outcome of an accident that has already occurred, such as eCALL. None of the systems investigated here had this attribute.

A new method was developed in this project to assess the mobility and comfort impacts of ITS systems on vulnerable road users. The method is analogous to the safety approach. The problematic aspect of applying the method was data availability, especially for comfort. The use of expertise, external as well as internal to the project to assess the impact, was a useful method to steer the assessment process. The methodology provides a good starting point for future work and methodological development.

The comfort and mobility impact assessment process was conducted for different road user groups, and for different age groups. Comfort has not been covered in previous studies to any great extent, especially not with the focus on vulnerable road users in relation to ITS. Therefore the assessment regarding comfort could not be based on evidence from literature or previous studies; the assessments had to be based on similar studies, if available, based on car drivers’ comfort in relation to ITS, or assessments from the experts within the consortium or the responses to the expert questionnaire. The concept of mobility is much more common in literature, but still the existing studies do not often cover mobility of vulnerable road users, and seldom in relation to new ITS.

Despite the limited availability of literature regarding mobility and comfort of vulnerable road users in relation to new ITS, much literature was reviewed to have as valid and reliable estimates as possible of the estimates of the effects. To increase the reliability of the estimates, several experts within the project contributed to the work in parallel and the estimates were cross-checked.

The eIMPACT methodology provided a systematic approach to cover all mobility and comfort effects of the systems; including not only the expected positive effects of the systems but also the indirect and negative effects.

The comfort benefits, the assessment of which was conducted in an experimental manner in this study, form a considerable part of the total vulnerable road user benefits for some systems, although safety benefits usually outweigh all other positive impacts [1]. This leads to the need for a standardised methodology and framework for vulnerable road user benefits assessment and especially for the assessment of comfort benefits. However, the proper application of an overall standardised methodology for benefits assessment requires the collection of harmonised safety and mobility data across EU countries, to be set at a level of detail sufficient to disaggregate system benefits not only to general vulnerable road user groups but also per age category.

Although not addressed in this study, the health benefits of active modes (walking and cycling) can have an impact on comfort levels and they are seen to outrun accident risks. ITS applications could be useful in improving the perception of these health benefits. Future analyses could take health benefits into account.

Because the systems analysed in VRUITS are not yet deployed, the possibility to validate the methods lies in the future. The safety impact assessment utilized methods described in detail in the literature. The mobility and comfort methodology was developed in this project. The VRUITS deliverables D3.1 [1] and D3.2 [16] make references to the relevant literature on approaches and formulae used. Chapter 2.3 of D3.1 [1] describes the procedure applied and clarifies calculations with examples. All assumptions and results used as input are made transparent for each ITS analysed (Chapters 4–12 of [1]) which makes it possible to compare the findings with earlier (and future) assessments. The transparency will enable the validation of the results in the future. The validation could be done in the future field operational tests focusing on measuring user behaviour, as it was done in DRIVE C2X project [11, 31]. In addition, new findings can be incorporated into the assessments. However, the detailed calculations were not included in this paper.

In order to improve the accuracy of the estimates, there is a need for better accident data (on number and details of accidents, including hospital records; especially regarding injuries), also to correct for injury underreporting. Also mobility data (such as quantitative detailed information about trip length, duration and frequency for vulnerable road users of different groups and age groups), and comfort data (such as how the users perceive their travel; i.e. change their opinion regarding the comfort of the travel undertaken, and perceived safety in relation to traffic) is needed. Finally, trials to test the functioning of the systems and their effect on road user behaviour are needed to better understand the effects of the systems.