Driver Assistance Systems in Oncoming Traffic Situations

In Germany, every fourth fatal road traffic accident takes place in situations with oncoming traffic. Two out of three fatal accidents occur on two-lane rural roads. Overtaking maneuvers and loss-of-control situations are responsible for many of these accidents and they usually result in serious injuries or fatalities. This paper analyzes the basic accident mechanisms in oncoming traffic collisions, focuses on human error that leads to the collisions, deduces target requirements for assistance systems, addresses safety benefits in terms of mitigating the severity of injury of occupants and vehicle damage of those involved. This paper presents the results of a driving simulator study that describes basic driver behavior in these situations. The paper also describes different variants of assistance systems that address these drivers behavior effectively by acoustic warnings. INTRODUCTION Worldwide the number of traffic fatalities has decreased in Japan, USA, Russia, European Union (EU), UK and Germany as shown in Figure 1. In its White Paper concerning the safety of road users, the EU sets as its common goal a reduction of 50 % in the number of fatalities among European road users by 2010. This EUinitiative has encouraged the introduction of more and more active safety measures as standard equipment or optional features in new cars. While in the past, systems for stability control and advanced brake assistance had been in the center of development efforts, the focus is shifting increasingly towards systems that can analyze environmental and situational conditions in complex traffic scenarios. They will increasingly contribute to an additional reduction of accidents. Using innovative sensor technologies and improvements in the area of situation analysis and assessment, even more complex traffic situations such as at intersections and involving oncoming traffic become usable for advanced driver assistance systems. In 2009, a total of 2.31 million traffic accidents were registered by the police in Germany. In these accidents 4,154 people were killed and another 397,671 were injured. At an 8 % margin, oncoming traffic accidents take a middle position in accidents causing injuries. However, they gain importance when considering accidents with fatalities or severe injuries. Here, oncoming traffic accidents account for 22 % of accidents involving fatalities and 17 % of accidents involving severe injuries. Observing accidents that happen on rural roads, but not on divided highways, this type of accident accounts for 32 % of all people killed; 774 out of 2452 fatalities occurred in rural areas [1]. According to official accident statistics from the US and Japan, oncoming traffic accidents account for 4 % or 3 %, respectively, of all accidents involving injuries (see Figure 2). IIHS [1b] reports that the amount of fatalities in accidents with oncoming traffic is nearly 24 % of all fatalities in road traffic. The percentages in the severe injury and fatality categories, however, are similar to German statistics. Russian authorities report 10 % oncoming traffic accidents with injuries [2] and approximately 33 % are fatal [3]. Dr. Schittenhelm 1 Figure 2: Distribution of the types of accidents involving personal injury. Figures for Germany (2009, DESTATIS), USA (2009, NHTSA) and Japan (2007, IATSS) Figure 1: Trend of fatalities in road accidents from 2001 to 2009 in Germany, Great Britain, EU(27), USA, Japan, Russia Currently, the research regarding oncoming traffic accidents has mostly considered aspects like road design and traffic theory and has focused less on the design of advanced driver assistance systems. For example, Wang et. al. [4] examined the estimation of conflict probabilities in overtaking situations. Hegeman et.al. [5] have analyzed the individual phases of the overtaking process and divided it into various sub tasks. Hohm et. al. [6] have researched possible approaches for an overtaking assistance system. As part of the PRORETA 2 research project, in 2009, Continental and the Technical University of Darmstadt presented a prototype of an assistance system supporting the driver while overtaking on country roads. The prototype shows that the technical implementation is possible [7]. This paper uses a different approach. Our starting point was not a technical implementation in a vehicle, but an examination of driver behavior associated with oncoming traffic accidents that resulted from overtaking another vehicle. Variables under considerations were the behavioral, attentional, perceptual, and psychomotor facets of driver behavior and performance. The detailed understanding of the mechanisms how these human factors interact and their sensitivity is necessary when designing an effective and user accepted assistance system in this specific pre-crash situation. The research was conducted in 2007 / 2009 in the Daimler AG (moving base) driving simulator in Berlin, effectively ruling out any risk to life or injury of the test persons. The experimental design was based on a detailed analysis of on-road accidents. In the first part of this study the human errors that lead to an accident were identified. Building on the test results, a second study was conducted analyzing the potential of a warning function and its user acceptance. The study also included a change in the test persons’ perspectives in the situation. In one instance the test persons took the “active” part as the driver in the oncoming lane of traffic. In a different scene the drivers were placed in a “passive” role in which another vehicle in the oncoming lane of traffic started an overtaking procedure into “their” lane, facing them directly. Both situations showed significantly different patterns of behavior. ACCIDENT MECHANISM AND RELEVANCE The pre-crash situations, which most frequently lead to oncoming traffic accidents were first analyzed. The analysis based on the representative GIDAS database, which will be introduced briefly. GIDAS database – a statistical representative sample of accidents for Germany The analysis in this paper is based on accident data provided by the GIDAS project. GIDAS is an abbreviation for “German In-Depth Accident Study”. It represents a cooperative project between the German Association for Automotive Technology Research (Forschungsvereinigung Automobiltechnik e.V., FAT) and the German Federal Highway Research Institute (Bundesanstalt für Straßenwesen, BASt) (see [8, 9] for more details). In its current form it was founded in 1999. Since then data for in-depth documentations of more than 2000 accidents per year were collected in two research areas – the metropolitan areas surrounding Hannover and Dresden (see Figure 3). The criteria for choice and collection are: (1) road accident, (2) accident in one of the research areas, (3) accident occurred when a team is on duty in a defined time frame, and (4) at least one person was injured in the accident, regardless of severity. For each accident a digital folder was created according to carefully defined guidelines and coded in a database. Depending on the type of accident, each case is described by a total of 500 to 3,000 variables, containing, e.g., accident type and environmental conditions (the type of road, number of lanes, width, surface, weather conditions, time of the day,...), surroundings of the accident scene, vehicle type, vehicle specifications (mass, power, tires, ...) and configurations (primary and secondary safety measures), documentation of damage to the vehicles, and injury data for all persons involved and their medical treatment. The investigation of all cases is “on the spot” to ensure the best visibility of traces for the best possible reconstruction. Each accident is reconstructed in detail including the pre-collision-phase. Available information includes the reconstructed initial vehicle and collision impact speed, deceleration, as well as the speed sequence of the collision. ACCIDENTS WITH ONCOMING VEHICLES Selection of accidents for detailed examination In the GIDAS database accidents are encoded according to the extended accident catalog of the GDV (German Insurance Association). The various accident types are derived from the situations from which the accident evolves. An oncoming traffic accident can be subdivided into the following five accident types: • Type A „Driving accident in a left turn.“ • Type B „Driving accident in a right turn.” • Type C „Driving accident on a straight road.“ • Type D „Accident in parallel traffic with oncoming vehicles.” • Type E „Accident in parallel traffic involving the overtaking vehicle and oncoming traffic” On the basis of about 1060 accidents belonging to type A to E selected from GIDAS 12-2007 it was found that Dr. Schittenhelm 2 Figure 3: GIDAS the research areas around Hannover and Dresden the subgroup of traffic accidents (type A-D), which ultimately lead to an oncoming traffic accident, had the largest share of about 60 %. About 25 % take place in curves. About 35 % of oncoming traffic accidents are preceded by a lane change (Type E). In the case of driving (or loss-of-control) accidents, with the ESC and the lane departure warning / protection / lane guiding system currently offered in the market, there are already assistance functions that address this accident type. The coming years will show how well these systems work in helping the driver to prevent these accidents. Accidents with oncoming traffic due to lane changes have so far not been addressed by a assistance system, which is why this type of accident was selected for being studied in the driving simulator. On the basis of 325 representative accidents (selected from GIDAS-2007) that were caused by “overtaking into oncoming traffic” these accidents can be characterized as follows: • The oncoming traffic accident preceded by a lane change is an accident that in 90 % of cases occurs on rural roads, usually well-developed trunk roads, typically with single carriageways. • About 60 % take place on (typically long) straights and about 35 % around the exit or after (typically shortly after the end of) curves. • At 6 %, it has an extremely high rate of fatalities. • Involved in these accidents are 80 % passenger cars, 15 % commercial vehicles and 15 % motorcycles. • Collision partners of the passenger cars are 70 % passenger cars, 17 % commercial vehicles and 13 % motorcycles. • Passenger cars collide at 45 % fully covered head-on, 10 % partially covered head-on and 15 % side-on while evading the oncoming vehicle. 15 % collide at the conclusion of the maneuver with the vehicle they have overtaken. • The driver of the overtaking vehicle overlooks the oncoming traffic or underestimates the distance required for the passing maneuver and/or the speed and its consequences. From this data it can be estimated that the oncoming traffic accident preceded by an overtaking maneuver has a share of about 8 % of fatalities on German road traffic. This result fits well with current figures of the Royal Society for the Prevention of Road Accidents [13] for UK. They conclude that in 2007 175 people were killed in overtaking (into oncoming traffic) accidents, with a further 1,351 seriously injured. This means that in the UK around 16 % of motorcyclist fatalities. about 6 % of all car occupant fatalities, and about of 7 % of all road fatalities occurred in this kind of accident. This GIDAS analysis was the basis for a representative routing and definition of the accident situation for the Daimler AG driving simulator experiment. Derivation of the experimental design The results of the study define requirements for the used test track and the scenarios for the experiment. Based on the results the goal was to create a test track that met the requirements for representative accident scenarios, thus the experimental design met these criteria: • The track passes over country and represents a well-developed trunk road. • The track has a long, easily manageable straight section that invites the driver to overtake vehicles • Before the "active overtaking maneuver" there is a long curvy stretch with dense oncoming traffic. At the beginning of the curves the participants approach a vehicle (M-Class) that drove through the curves at about 100km/h. In the curves there is a speed limit of 100km/h. • As in reality, traffic is simulated at irregular intervals on the entire stretch. • The stretch has a length of 70 kilometers. • The driver repeatedly experienced harmless scenes in order to convey a natural driving sensation. The individual events were evenly distributed over the entire stretch. On the drive they repeatedly went through sections with and without a lead vehicle. • The participants repeatedly experienced overtaking in oncoming traffic. They were also able to overtake several times on their own. • The order of "active" and "passive" is selected at random for each participant at the start. • The participant is seated in an vehicle cabin of a C-Class with an automatic transmission. SIMULATOR Driving simulators are suitable especially in the early phases of system design for safe and repeatable tests of the interaction between “normal driver” and primary safety measures in critical situations. Results obtained with this method have the advantages over others because they offer a high degree of determinateness, reliability, objectivity, validity and therefore transferability for instance in different set ups and comparability for example between different levels of development. On the other hand, there are a few drawbacks, such as extraordinary expense for hardware, software and operation, integrated simulation chain in the design process, as well as specific difficulties. for example in replicating the vehicle movements, graphical presentation and limited awareness of exposure. The mechanical set-up of the Daimler moving base driving simulator in Berlin presented in Figure 4 and is described in detail in Käding [12]. This well established simulator provides a very realistic driving environment. The movement system is composed of a hexapod and a cross cylinder. It allows for a movement of ± 3.80 m in transverse direction and of 1.50 m in longitudinal direction. The dome includes a CRT-projection system of 230° to the front, 60° to the back and exchangeable standard vehicle. LCD displays were integrated in the side mirrors of the test vehicles.