UR:BAN Human Factors in Traffic

The automobile and road traffic have made in the course of their joint development great progress. Still, to be ready for the future, the entire transport system has continuously to meet new and higher requirements. Each trip shall be traveled safely, efficiently and comfortably. The progressive urbanization leads to more and more people living in urban areas. The additional large number of commuters and supply transport should not be underestimated. This moves mobility spaces increasingly into the urban space and individual traffic using cars represents the majority.

Heterogeneous user requirements lead to conflicts and reduced efficiency in a limited urban space. In industrialized countries such as Germany 85% of the population live in the urban area. While especially outside Europe but partly also in Southern Europe a clear trend toward megacities with several million inhabitants can be observed, urbanization in Germany happens in a variety of medium‐sized cities, as well as some major cities. These agglomerations are adjacent to a core city and reach out to a far area. They are characterized by strong commuter flows with mixed traffic (car/motorcycle/bicycle, passengers of public transport, pedestrians, etc.). This leads to increasing mobility needs in the urban context and suggests to focus research activities towards an efficient urban mobility.

As urban traffic is fundamentally complex and multi-faceted, so are inevitably any research initiatives that aim to investigate associated research questions thoroughly and comprehensibly. In order to provide a superordinate structure for these research efforts, on the basis of which individual research activities can be coordinated and synthesised, the project Urban Driving (abbrev. UF for “Urbanes Fahren”) was initiated. By defining a common terminology, data formats and research methodology for the different research activities within UR:BAN, the multiple results across the UR:BAN research initiative could be aggregated and summarised in form of core messages. The following chapter provides an overview of the activities and key results of the project UF, thus introducing a meta-perspective of urban traffic research.

In the German national funded UR:BAN project numerous assistance functions supporting the driver in longitudinal and lateral vehicle guidance control up to preventing accidents by warning the driver or making evasive manoeuvres autonomously are considered in order to contribute to improve safe, comfortable, and energy‐efficient driving in urban areas. Especially, the design of the human‐machine‐interfaces (HMIs) of all the applications together faces many challenges. For example, urban traffic is compared to highways and rural roads more complex and dynamic due to different infrastructure and the interaction with other road users (e. g. crossing vehicles, pedestrians, and bicyclists). Thus, information, warning, and system intervention concepts should be designed to minimize the conflict between the complexity of urban traffic situations and the driver’s limited cognitive information processing by keeping understanding and trust in such assistance systems. The subproject “Human‐Machine Interaction for Urban Environments” (named “Stadtgerechte Mensch‐Maschine‐Interaktion”) addresses the development and design of HMIs of this range of driver assistance systems. The central goal of the subproject is to design user‐oriented, integrative HMI concepts of current and future assistance systems by considering the challenges and needs in urban areas. In total, nine project partners – Adam Opel AG, AUDI AG, Robert Bosch GmbH, Daimler AG, MAN Truck & Bus AG, Technische Universität Braunschweig, Technical University of Munich, Universität der Bundeswehr München, and Volkswagen AG – rise to this challenge.

The development of driver assistance systems has shown various innovation waves in the last few decades. While such systems can contribute to increasing driving safety and comfort on highways and rural roads, future systems will be enhanced to include urban traffic. Such trends also face challenges, especially in the design of human-machine-interfaces (HMIs). In the UR:BAN project, one aim was to design user-oriented, integrative HMI concepts of current and future assistance systems by considering the challenges of urban driving. Therefore, the cross-functional “HMI tool kit” was developed comprising a strategy for the systematic derivation of action-oriented HMI concepts. The structure of the HMI tool kit differentiates between cases of applications for safe, comfortable, and energy-efficient driving, which takes requirements of different time horizon and user-actions into account. The HMI tool kit provides insight into 1) how information, warnings, and system interventions should be filtered and prioritized to the driver concerning the current traffic situation and 2) how the preferred driver behaviour can be achieved in the situation by the selection of suitable HMI components (e.g. display in the instrument cluster, warnings or sounds). The aim is to warn the driver adequately in safety-critical situations and allow him to develop a sufficient understanding for continuous system interventions or advanced navigation recommendations. The UR:BAN HMI concepts can contribute to an anticipatory driving style, to mitigate safety-critical situations, and to improve stress-free and low-emission driving in urban traffic.

One of the most important aims of driver assistance systems is the prevention of accidents. Onboard-sensors, algorithms, and other technologies allow developing strategies to achieve this, for example by warning drivers in critical situations. Depending on the remaining time, different driver reactions may have to be elicited. In some situations it may be sufficient to slow down well in advance or to change lanes. However, in other situations only an emergency braking or a fast evasive manoeuvre can prevent a collision. Since in most situations drivers are still in control of the car, the question arises of how to best support them in these kinds of situations. This encompasses two basic HMI aspects: (1) How can the required reactions be elicited in drivers, before the system has to intervene automatically due to the increased criticality and reduced time left? In such a very critical situation, (2) how can drivers be explained how and why the assistance system has taken over and intervened? Within the scope of the UR:BAN project, HMI concepts were developed and evaluated with regard to these aspects of intervention and warning strategies. The chapter gives an overview about the conducted studies and resulting HMI concepts.

Advanced driver assistance systems (ADAS) which continuously intervene in the lateral or longitudinal control of the vehicle can increase the efficiency and comfort while driving. In order to achieve the best possible use of such systems, the HMI has to be adapted to the driver’s needs and capabilities. The current chapter describes how two HMI concepts for urban traffic were developed and refined based on the driver’s demands: (1) a truck-specific HMI strategy for an automated longitudinal control and (2) a HMI strategy for a lateral control system on narrow roads – the constriction assistant. In both cases, the essential components of the HMI were identified using several methods, e.g. accompanying truck drivers in daily traffic, driving simulator studies, and integrated into an HMI strategy.

Recommending assistance systems provide driving recommendations, with the goal to improve driving comfort and/or efficiency. The vehicle sensor system integrates information that is not available for the human drivers. The drivers benefit from the provided information because it increases their knowledge on the current and upcoming driving situation, which allows adaption to the driving behaviour. The chapter provides two empirical approaches to recommending assistance systems as presented in the UR:BAN project. The first part of the chapter includes the development of the HMI strategy of a traffic light assistant as an example for a recommending driver assistance system. The studies covered the development of the visual HMI concept and the investigation of the influence of platoon driving and complex traffic conditions on the evaluation of the assistant. The second part presents studies for the development of a generic, integrative HMI concept for different recommending assistance systems, considering a multimodal selection of HMI components (such as HUD, instrument cluster, and force feedback pedal).

The recognition of driver intentions can enhance the driver-vehicle interaction and offer more intuitive assistance and driving support. The assistance should display comfortably timed warnings only in situations where they are really needed and has to act in accordance to the driver’s intentions. This was the framework of the UR:BAN sub-project VIE on detecting driver’s intention and predicting his behaviour. This introductory chapter starts with a description of the project’ objectives. It gives an overview on the working method during the project and summarises the results. Finally, it gives some indications on the main topics of the following chapters presenting parts of the project’s activities in detail.

The analysis of real world driving data entails numerous challenges. In this chapter, several strategies are proposed to meet challenges that surface in data storage, data extraction, data correction and data enrichment. The strategies are illustrated with examples from a study that had been conducted as part of the UR:BAN research project ”Behaviour Prediction and Intention Detection” (VIE), which aimed at investigating driving behaviour when approaching intersections under real environmental conditions in order to predict turning manoeuvres at urban intersections. It was demonstrated that with the proposed data infrastructure, correction procedures and extracted filters for potentially confounding variables, it is possible to establish a “clean” data basis to implement and adjust a prediction algorithm for turning manoeuvres according to individual driver characteristics.

Driver distraction strongly influences the accident risk on both motorways and urban streets. In this context, visual distraction - long glances off the road - is the most contributing factor. However, in natural driving engagement in visually distracting activities is very frequent compared to the small number of critical incidents. This indicates that drivers apply situational-adaptive gaze and driving strategies that can provide a certain amount of driving safety. Yet, most state-of-the-art mitigation systems assess driver distraction based on fixed thresholds on glance duration. This chapter presents an approach for prediction of situation specific human behaviour in distracted driving. Here, we apply a driver model based on sub-optimal control. Taking into account driver strategies and their potential insufficiencies in the current driving context, our method has the potential to greatly improve assistance systems, by reducing unneeded warnings and interventions. This holds true especially in urban scenarios that are characterized by a broad variety of driving situations.

Lane change manoeuvres pose high demands on the driver. Driver intent information is supposed to provide lane change assistance specifically when required, thus increasing acceptance and traffic safety. Based on an on-road study including 60 participants, the I-FAS investigated lane change behaviour at different levels of analysis. The present chapter shows the analysis of lane change predictors on the behavioural, strategic, manoeuvring and control level. Considering driver characteristics on the strategic/behaviour level, familiarity with the route resulted as the most important predictor for the number of lane changes performed per trip. Analyses at the manoeuvring level showed that lane change manoeuvres need to be further subdivided into subtypes with different requirements for prediction. Driver behaviour – especially automated glance behaviour at the control level – differed considerably between e.g. lane changes due to a slower vehicle ahead and lane changes on an added lane. Mirror glance patterns for specific lane change types resulted as promising and quite stable intention predictors, even before the activation of the turn signal. However, the interpretation of glances as indicator for lane change intention is vague without the integration of information about the driving situation. Therefore, a realtime lane change prediction algorithm was developed integrating driver behaviour, vehicle parameters as well as data from the vehicles’ surroundings in a Bayesian Network.

The task of driving is very complex and highly demanding for the individual. The optimal driver assistance strongly depends on the situation and the driver’s needs. In particular, this applies to driving manoeuvres as lane changes. Consequently, future advanced driver assistance systems will have to detect and assess driving situations as well as the driver’s intentions automatically before a driving manoeuvre is initiated.

The method proposed predicts situations of upcoming lane changes based on assessments of the environmental situation and the driver’s behaviour. For this purpose, information gained from a 360° sensory perception of the vehicle surroundings and from the analysis of the driver’s gaze behaviour is fused by means of a Bayesian network. The implemented algorithms work in real-time and provide a probabilistic estimation of the intention of the driver to perform a specific manoeuvre. The application of prediction was integrated into a test vehicle and evaluated by using real traffic data and driving studies.

Advanced driver assistance systems (ADAS) can help to reduce road accidents. In order to increase the positive effects of these systems the warnings and interventions have to be adjusted to the driver’s behaviour. With regard to the human behaviour a real time capable algorithm for predicting driver’s manoeuvres was developed.

The basis for development of the algorithm were two controlled field studies with the focus on inter- and intraindividual behaviour in urban traffic situations. The algorithm is based on Fuzzy Logic and Edit Distance; it was trained with the data containing the driver’s behaviour before and during driving manoeuvres of the field study. Focus lies on the driver’s vehicle control and his head- and gaze behaviour. The feature selection was performed on the basis of true and false positives, false negatives and the prediction time horizon. The algorithm learns behavioural sequences of considered features in an offline training step and compares the actual driver’s behaviour with the trained sequences during the real-time detection to calculate the manoeuvre probability and the time horizon until a certain manoeuvre will be conducted.

Driver intention recognition can enhance the driver-vehicle interaction by offering more intuitive assistance and automated driving support. Especially urban environments require fast reactions and hence assistance systems which act in accordance to driver’s intentions. Assistance should provide comfortably timed warnings only in situations when drivers really need this support and not in situations when the driver is already intending to react to a thread.

Fraunhofer IAO developed an algorithm in the UR:BAN MV subproject VIE to detect driver’s intention to brake when passing a pedestrian. The Fraunhofer algorithm analyses eye gaze data in correspondence with pedal activity to judge the driver’s attention on the pedestrian and the readiness to brake. BMW implemented the algorithm in the UR:BAN KA subproject SVT in a research vehicle and combined it with an environmental analysis of the situation.

In a test scenario the timing of a warning to the driver was adapted to the recognized intention to brake. Together with BMW’s pedestrian intention recognition algorithm, the driver intention recognition allows early warnings, while limiting the frequency of warnings to really relevant situations.

The sub-project Simulation (SIM), which lies within the project pillar Human Factors in Traffic, focuses on the analysis and descriptive modelling of the behaviour of individual road users and their interactions with one another. This is done in consideration of newly developed driver assistance systems and intelligent transportation systems. The objective of the sub-project is to improve and extend driving simulators and microscopic traffic simulation in order to simulate and study the resulting behaviour of the road users in a more realistic way. The driver-vehicle-environment system is investigated in detail using many different research environments. At a controlled test site, a number of relevant traffic scenarios are observed with a particular focus on urban traffic and interactions between motor vehicles and pedestrians and bicyclists. Interactions between multiple real road users are analysed using connected driving simulators. Microscopic traffic simulation is used to study urban areas with a large number of simulated road users. In this environment, a particular focus is placed on improving the modelling of bicyclists and pedestrians.

As one part of the subproject SIM pedestrian-vehicle interactions were investigated based on three different methodical approaches. The conducted studies focused on the pedestrian crossing intention in case of an approaching vehicle. The three approaches for data collection were chosen with the goal of modelling the pedestrian crossing probability depending on situation-specific parameters. In the first approach, subjects drove an equipped vehicle under real traffic conditions on a route which is highly frequented by pedestrians without knowing the actual objective of the study. The behaviour of the interacting pedestrian as well as the driver behaviour was investigated based on the objective data from the environment sensors and the driver input. Based on these results the second approach was set up as a study in a controlled environment in which the behaviour of the drivers (confidantes) was systematically varied. In addition, different vehicle types (passenger car, truck) were considered. Both approaches were supplemented by static traffic observations (third approach). The results show a situational influence through e.g. crossing aids and type of interacting partner (passenger car, truck) within the previously described pedestrian crossing events. The modelled crossing probabilities are in line with the results of the linked vehicle-pedestrian simulations of the subproject SIM and serve as a valuable input for the interaction detection and the behaviour modelling by helping to understand the natural interaction behaviour of pedestrians during crossing scenarios.

Especially in urban areas, more people choose the bicycle over the car in order to get from point A to point B. On the downside, more bicyclist in the streets resulted in a 50 percent increase of fatalities. Last year, in Germany, bicyclist accounted for twelve percent of all fatalities. Research shows that a major contributing factor to those fatalities is ‘insufficient cooperation’ between bicyclists and motorists. Nonetheless, most encounters of bicyclists and motorists neither result in conflicts nor in fatal crashes. In order to investigate interaction patterns between bicyclists and motorists, those road users were observed at a busy intersection in Braunschweig for a period of five working days. Situations, recorded at the AIM research intersection, in which motorists turned right and bicyclists went straight through the intersection, were analyzed. The goal was to understand the behavior and the underlying mechanism, quantify the observed behavior, and identify objective parameters to map the behavior. As a result, the knowledge may be used to implement strategies and technologies that may predict and prevent fatal crashes in intersections.

Microscopic traffic simulation tools, which are often used to evaluate road infrastructure design, signal control, intelligent transportation systems (ITS) and advanced driver assistance systems (ADAS), are limited in their capacity to realistically simulate bicycle traffic. Due to their small size and high manoeuvrability, bicyclists are one of the most flexible road user groups. They are faced with tactical choices that do not apply to other road users, such as the choice between using a bicycle lane, the roadway or the sidewalk, the option of travel with or against the mandatory direction of travel and the selection between multiple possibilities for executing a left turn. This flexibility makes it important to understand and accurately simulate not only the operational behaviour, such as speed and acceleration, but also the tactical decisions of bicyclists. In this chapter, the data collection, data processing, behavioural analyses and model development carried out within the project Urban Space: User oriented assistance systems and network management (UR:BAN) are presented. Analyses and modelling approaches for three aspects of the operational behaviour of bicyclists are presented in detail; speed, acceleration and spacing. The methods used to analyse and model the tactical behaviour of bicyclists are briefly introduced.

Interaction between human drivers and programmed agents under specific experimental conditions is the most common approach in driving simulation research nowadays. Social interaction, especially in transient, complex situations such as those occurring in urban traffic, is a multi-facetted, multidirectional, and above all dynamic phenomenon. The programmable traffic participants (bots) in simulations may thus encounter interaction constraints. This chapter describes the UR:BAN approach, where the narrow spectrum of human-bot interaction is expanded. The apparatus consists of a multiparty simulator. A car driver in a driving simulator encounters a pedestrian in a second simulator in varying situations. Both meet in the same simulated environment and are able to mutually adapt their behaviors relative to the other road user. Traditional safety-oriented data analysis, involving for example time to arrival, post-encroachment time, and deceleration-to-safety time, was conducted and supported by novel approaches using cross correlation, cross recurrence quantification analysis, frequency analysis, sense of presence, and immersion assessment to take the features of social interaction into account.

Cooperative intelligent transport systems (C-ITS) are among the most promising innovations in present automobile industry. When evaluating those systems, possible effects on surrounding traffic have not been investigated properly. The necessity to research encounters of drivers with and without such systems is discussed. Tools, methods and parameters to enable researching these encounters are outlined and argued in terms of their advantages and disadvantages. First study results focusing on encounters of unequipped vehicles’ drivers (UVDs) with a driver with traffic light assistance system are presented. The study results and their implications highlight the need for system developers and researchers to take effects on UVDs into account for the development and evaluation of C-ITS. Remaining unanswered questions are presented to pave the way for new research initiatives.

Driving simulations and traffic simulations are used successfully for research in traffic sciences. However, these tools do not allow the investigation of social interactions between drivers. These are always present when two or more drivers encounter each other, e.g. during car-following, crossing, merging or oncoming traffic situations. In contrast, the connection of multiple driving simulators enables the investigation of social interactions. The result is a so-called multi-driver simulation which consists of several driving stations that are used by the participants to drive through the same virtual and controlled environment. In the virtual environment, the drivers are able to see the other vehicles and can react to the other participants’ behavior. By using empirical data from own studies this chapter shows the additional value of the multi-driver simulation compared to the single-driving simulation with one driver and to the traffic simulation.

The following chapter focusses on the interaction between motorcyclists and passenger car drivers. Accident statistics reveal the need to deal with this topic. Hence, the within UR:BAN SIM chosen approach of building up a Motorcycle-Car Multi-Driver Simulation is described. The technical setup, challenges and advantages as well as study examples will be addressed.

Since a few years, multi-road user simulators offer the possibility to investigate social interactions in virtual study environments. However, this new tool requires an own study methodology. Therefore, the aim of this chapter is an encompassing methodological consideration of multi-road user simulations. The methodological conclusions of several studies using various multi-road user simulations (driver-driver simulation, multi-driver simulation, driver-pedestrian-simulation, driver-motorcyclist simulation) during the UR:BAN project are summarized and demands and recommendations concerning study planning, study conduction and data analysis are presented. The benefits and challenges of multi-road user simulations are compared to single-driver simulations. Finally, it is considered for which research questions the multi-road user simulation is appropriate.

The subproject UR:BAN KON “Controllability“ developed and evaluated methods which can be used to assess safety and controllability in early-stage development of new driver assistance systems. The empirical studies focused on emergency steering and evasion assistants that help the driver to avoid collisions in time-critical scenarios. Several factors like available manoeuvring space, drivers’ attention and characteristics of the system design could be identified which influence controllability and safety in use. Additionally, results from several research environments were compared and evaluated regarding their validity.

According to standards such as ISO 26262, system controllability for human drivers must be ensured with new driver assistance systems. The choice of an appropriate research environment is a central issue in controllability research and precedes questions such as the criticality of test scenarios and deduction of pass-fail-criteria. The methodological trade-off between research on test tracks and with driving simulators cannot be resolved easily. Although test track research allows for the analysis of human interactions with real vehicle dynamics, the investigation of critical situations with other traffic agents requires considerable effort and is sometimes not feasible. For example, the complexities of real-life urban scenarios cannot be readily replicated on test tracks. Driving simulations do not underlie these restrictions, but limitations concerning visual and proprioceptive feedback raise questions of validity. Within the UR:BAN MV KON project, a study was performed on criticality perceptions towards various metrics (longitudinal and lateral distances and decelerations) across four research environments. A total of five experiments were performed using a dynamic driving simulator, a static driving simulator, a test track vehicle, and a Vehicle-In-the-Loop (VIL), which is a hybrid between a test track vehicle and a driving simulator. In the present chapter, we present results from the series of experiments and show how the results relate to existing validity research.

Emergency steering interventions are the progression of existing emergency braking functions and subject to current research initiatives like UR:BAN (Manstetten et al., 2013) [1]. Such systems are designed to potentially prohibit accidents, even if it is too late to brake to a standstill. Otherwise, such interventions with a lateral component introduce new challenges to controllability assessment. The Vehicle in the Loop (VIL) is a combination of a test track vehicle and a driving simulation offering new methodological access to controllability aspects (Berg & Färber, 2015) [2].

Two studies are discussed in which controllability aspects of emergency steering systems are investigated with the VIL. The driver-system interaction at emergency steering interventions is subject of the first investigation. Especially when system interventions could cause collisions with obstacles on the opposite lane, drivers can adjust the steering interventions appropriately to the driving situation. These results were observed in both test track vehicle and VIL, which suggests the VIL to be a valid testing environment for the investigation of emergency ADAS. The second experiment considers aspects of emergency steering systems from the perspective of opposing traffic. This approach shows the collision avoidance potential of other road users with automatic steering interventions that are not appropriate to the driving situation and provides clear limitations for the functional design of such systems.

One of the goals of the subproject KON was the investigation of the applicability of existing controllability evaluation methods of assistance systems operating in urban areas. Due to the use of these systems in urban traffic conditions different boundary conditions have to be considered with regard to the space available for specific driving manoeuvres by the driver. A real-vehicle study was conducted which addressed the influence of the available evading space on the driver reaction to steering interventions for evading assistance. The study investigated the reaction to system-initiated interventions for varying available evading space during normal system use as well as false activations. The data analysis showed that drivers adapt their reaction: They allowed a higher lateral deviation if evading space is available and suppressed the system intervention if the available evading space is limited. This should be considered for possible enhancements of the evaluation methodology of steering interventions. While existing controllability criteria which are based on the exceedance of the own driving lane seem therefore not to be applicable for the investigated system interventions, the driver adapt their reaction to the available evading space.

The development of evasion systems is a challenge when it comes to the design of the human machine interaction. To effectively assist the driver and prevent a collision, an emergency evasive manoeuvre has to be highly dynamic which may have an adverse effect on controllability in case of system failures. Therefore it is important to find a good trade-off between effectiveness, safety in use and controllability when designing such a system. There exist several different possibilities to perform an emergency evasive manoeuvre. At the moment, systems using directional torque overlays to perform an evasive manoeuvre are favoured although systems using differential braking, steer-by-wire or even combinations are also discussed.

Experimental studies have already shown that the characteristics of a steering torque overlay influence controllability as well as safety in use. However these results cannot be easily transferred to the design of emergency steering and evasion assistants because a design based on these results might render the system ineffective. This chapter summarizes current research activities and the resulting implications regarding the design of emergency steering and evasion assistance systems with a focus on driver-system interaction.

Executing secondary tasks while driving can affect situation awareness, reaction times and response selection by the driver. This may result in a decreased performance of the driver’s control in case of sudden system interventions, e.g. of an intervening emergency function due to a false activation. To date a sufficient literature base concerning systematic implementations of driver distraction in controllability tests is missing. The aim of this chapter is to summarise the methodology used in driver distraction research and to give recommendations which methods could be used in controllability research. Additionally, we will demonstrate the effect of driver distraction on controllability in two exemplary empiric evaluations.