REAL-WORLD CAR ACCIDENT RECONSTRUCTION METHODS FOR CRASH AVOIDANCE SYSTEM RESEARCH

Seoul 2000 FISITA World Automotive Congress June 12-15, 2000, Seoul, Korea F2000G317 REAL-WORLD CAR ACCIDENT RECONSTRUCTION METHODS FOR CRASH AVOIDANCE SYSTEM RESEARCH Thierry Hermitte 1) Christian Thomas 2) Yves Page 1) Thierry Perron 2) 1) Centre Européen d Etudes de Sécurité et d Analyse des Risques, 132 rue des Suisses 92000 Nanterre, France 2) Laboratoire d Accidentologie de Biomécanique et d études du Comportement humain PSA PEUGEOT CITROËN RENAULT, 132 rue des Suisses 92000 Nanterre, France Development of crash avoidance systems and active safety systems must not be only based on experimental knowledge. The goal is to provide an efficient answer to still unsolved severe real-world car crashes which occur despite enhanced passive safety devices. This requires to know precisely the pre-crash conditions during about 3 to 10 seconds before impact. The paper describes the multidisciplinary systemic approach leading to the comprehensive methodology used in accident reconstruction in order to determine the best scenario, and to assess initial car speeds, paths and events in the different phases of the accident. This has already been carried out for about 400 car crashes with car occupant injuries (including 6% fatal and 10% severely injured). The necessity of collecting data on the spot of the crash scene is highlighted. Three well-trained investigators are involved. The first technician collects vehicle characteristics which could change later like tire pressures or gear position. The second technician focuses on measurements of road-marks and skidmarks, accurate impact point location, final car positions and weather conditions as well. The third technician is a psychologist who makes driver interviews in order to record immediately the story of the pre-crash (perceptions, interpretations and actions). After complementary data collection later on, the mathematical computer reconstruction can start taking into account behavioral driver activities. The basic mechanical equations are used to obtain primarily reliable scenarios for post-crash, crash and pre-crash phases, fitting together geometrical data and testimonies. Some kinematics parameters are estimated by a range of probable values with the Monte-Carlo method. Upon this first step, a more accurate accident simulation can be calculated with a reconstruction software. Results and examples of applications are given in the most common pre-crash situations (curves, straight roads and intersections). Keywords: Accident Reconstruction, Accident Analysis, collision, primary safety INTRODUCTION In-depth accident investigations can be divided into 2 subactivities : 1. The first one deals with secondary safety. The goal is to understand the injury mechanisms in real-world accidents and to improve occupant safety in cars by the means of protection devices or car structure. Almost all car manufacturers all over the world and even public research institutes have been carrying out that kind of study for decades. Specially trained accidentologists collect relevant information about car deformations and occupant injuries and feed it into a corresponding database. They don t need to go on the scene of the accident. Information is collected by accidentologists a few days or a few weeks after the accident at hospitals and at wreck garages. This methodology leads to a wide range of researches estimating risk curves or evaluating the effectiveness of on-board protection devices. See for example [10]. 2. The second one deals with primary safety. Car manufacturers started this activity in the early nineties, when it appeared that secondary safety would necessarily have limits and that there was a need for crash avoidance as well as a need for occupant protection. The challenge in this field is to understand the accident process, purpose new functions for active safety systems, and eventually to evaluate the effectiveness of new safety devices or avoidance systems on any kind of motorized vehicles. Previous, and even old, researches clearly stated that the best way to understand the causation process was to go on the scene of the accident as soon as the accident occurred [11], [12]. Prospective investigations help in collecting as much information as possible to be analyzed later on by using a relevant accident analysis theoretical and empirical model. Actually, later investigations, several hours or even several days after the accident, prevent the analysts from getting useful information such as the interview of the occupants, the witnesses evidence, the road state, the vehicle skid marks on the road, the condition of traffic, and so on. It is true that part of information can be found in police reports but experience shows that police reports do not provide indepth information. As an accident is clearly the result of inadequate interactions between the driver behavior, the road infrastructure configurations before and at the scene of the accident, the vehicle dynamics and state, and of course the conditions of trip (weather, traffic, passengers, etc.), accidents need to be investigated by a multi-disciplinary team, composed by a psychologist, a road engineer, and a vehicle engineer. Contacts and agreements have to be concluded with the local police authorities so that the team can be called up whenever an accident occurs in the investigation area. Agreements must also be signed with the ministry of Justice to allow that kind of technical work on accidents apart from judicial process involving drivers at fault. Investigations are exclusively technical and are carried out for research purposes only. In France, three institutes are presently carrying out that kind of in depth investigations with regards to primary 1

safety concerns: the National Research Institute for Transport and Safety (INRETS) and The European Center for Safety Studies and Risk Analysis (CEESAR) with The Laboratory of Accidentology and Biomechanics (LAB). Up to now, the two teams at CEESAR and LAB carried out about 600 accident investigations (mainly car to car accidents or single car accident just a few accident cases involved a pedestrian, a two-wheeler or a truck). The investigation area is very limited to a radius of 30 kilometers around the team s office. They collect directly the information if it is available on the spot or later on, at the hospital. They seldom use the police report. If they are called by the rescue service and come on the accident scene after the driver left the scene or the vehicles were moved from their position of rest and there is not clue to find the crash point and the place of vehicles at rest after the crash, the accident is not investigated. The objective is really to get as much information as possible to be able to make a cognitive and kinematic reconstruction of the accident. INVESTIGATION METHODS The collection of the information on-the-spot takes about two hours. Complementary collection is made afterwards at the hospital (second interview of the involved person especially and collection of the injury form) or at the local transport authority to get further details about the traffic or the road configurations. Most of the data is then coded and filled in a special database. Information that can not be coded is conserved in original dockets along with photos, sketches and sometimes video movies. On one hand, the way to proceed such accident investigations has necessarily a lot of advantages. It is the best way to understand the accident because teams come on it with the police and the rescue services and has the opportunity to collect information and testimonies on the spot. They are then close to the accident and their growing experience provides high level knowledge about accident causation. It will never be possible to know more about what happens in an accident without being in the car with the driver and reading in his mind while he is putting himself in an emergency situation. On the other hand, this investigation methodology has drawbacks. For instance, it does not allow to directly estimate the national statistical prevalence or the incidence of a risk factor just because CEESAR does not investigate a lot of accidents (120 per year while the official statistics count 120 000 personal injury accidents in France), because investigated accidents are fatal (fatal accidents are therefore under-represented), and because accidents occurring around the two investigation areas can not be considered representative of all accidents occurring all over France. However large databases do not record as much parameters and they are often biased towards severe crashes. The outcomes of such accident investigations are a complete database and accident researches that help car manufacturers to understand the accident genesis, and noticeably the car collision course and the drivers actions or failures in emergency sub-phases, and to evaluate a priori situations for which driver assistance or safety devices could have avoided the crash or slighted its injury consequences. RECONSTRUCTION The analysis of the accident starts with a simple question: what happened? The accident has to be reconstructed to understand its mechanism. Reconstruction does not only mean to estimate the driving speed of the involved vehicles, because this parameter is not enough to explain the accident itself. Crash speeds just give a clue about the violence of the crash : it can explain the occupants injuries or car deformations, but gives no information on the car motion and the driver behavior just before the crash. Driving speeds at the beginning of the conflict zone (i.e. the zone where the driving situation turns out into an accident situation) are also not sufficient to determine what could have been the active safety systems which could have avoided the collision or reduced its consequences. Carrying out an accident reconstruction can be compared to a crime inquiry where the detective is the accident expert. Putting together and selecting relevant information collected on the scene in time or/and afterwards allows a reconstruction and lead to a probable scenario of the accident : the what happened question becomes answerable. Actually, the reconstruction must argue about the following questions : What? When? Who? Why? How? First, the accident is divided in 4 phases: the driving phase, the pre-crash phase starting with an event precipitating the former situation into an accident situation, the crash itself, and the post-crash. Sub-phases can be identified inside any phases. It begins 10 to 15 seconds before the crash and ends with the vehicles in their final position after the crash. Reconstruction provides the relationship between the kinematics parameters (collision course, speed, acceleration, etc.), the driver behavior and failures (perception, evaluation, interpretation, decisions, actions, etc.) and the typology of the road (profile, grade, friction coefficient, surface, geometry, etc.). The pre-crash phase starts from the beginning of the conflict zone and simulates the behavior of each vehicle up to the first collision between them. It can be simulated independently of each vehicles. The crash phase describes the deformation step during the collision. This phase is very brief, only some milliseconds duration. The last phase is the post-crash phase. It deals with the course of each vehicle from the collision out to the position of rest. It can be studied for each vehicle separately. All these phases are linked by the speed : the speed at the end of a phase is the speed at the beginning of the next phase. With respect to pre-collision, a pre-crash table was especially conceived to analyze road accidents. This table shows, by time steps, the evolution of kinematic parameters about the vehicle (speed, deceleration or acceleration, distance from the Point of Impact (POI), lateral position from the right road side) and the cognitive parameters about the driver (control, visibility, perception, interpretation, action). To some extent, it can be considered as a story or as an abstract of the accident (Table 1). The starting point is the POI. The speed and the distance of the car to the POI can be calculated by applying a deceleration at any time. Control, visibility and perception are given a Yes or No value for each time step depending on whether the driver could control his vehicle, see or perceive the obstacle. Interpretation means whether the driver felt himself in a safety, risk or danger situation. Action provides the action of the driver (brake, throttleoff, accelerate, continue, turn right, turn left, ) in regard to the kinematics parameters. Cognitive information are 2

usually analyzed using interviews on the scene with the involved drivers. An example is given in the case study. Table 1 : Pre-Crash table Impact Sub Phase 1 Sub - Phase 2... Sub - Phase i... Sub-Phase p Impact Sub Phase 1 Sub - Phase 2... Sub - Phase i... Sub-Phase p Time Time Acceleration (m/s 2 ) Relevant behaviour Quantitative Parameters Distance to crash point (m) Speed (km/h) Transversal position in lane (m) Slip angle ( ) Behavioural Parameters Control Visibility Perception Interpretation Action A collision is defined by a crash between a vehicle and another vehicle, a pedestrian or a fixed obstacle. A single vehicle accident with no obstacle involved is a crash without collision. In the case of multiple collisions, each collision is considered one by one, collision by collision, with for every single collision, only two vehicles. The methodology of reconstruction follows 3 main steps: Analysis of crash configuration Rough calculation Simulation of the accident ANALYSIS OF CRASH CONFIGURATION The crash configuration gives the positions and the orientations of the vehicles at the impact. They can be determined by the observation of the vehicle deformations. The location of the point of impact (POI) is the main aspect of this step. Sometimes it is easily found by inspecting the road or by eye witness testimonies, but in general the POI is not obvious. Its location is set upon the elements collected on the accident site such as tire braking or skid marks. Techniques of determination of the POI depend on the crash types (single vehicle accident with loss of control, head-on collision, rear-end collision, frontto-side collision, ). Let us give 2 examples: - In case of collision at a junction, the braking marks left by the right of way vehicle tires are straight and start in another direction after the impact (braking or skid mark). The point of impact is at the exact location where the investigator finds a break in the direction of braking marks. - In case of head-on collision, where the first vehicle kept on going straight forward and the second one, such as pedestrian or motorcyclist was pushed straight backward (i.e. post-crash directions are parallel) diagrams can be used along with the sketch. Those diagrams (Figures 1 and 2), provide the relation between throwing distance from the POI to the position of rest of pedestrian or motorcyclist and the collision speed of the encountered car, or the relation between the collision speed and the brake distance of the car, allow to estimate the POI when they are juxtaposed. The knowledge of the deceleration coefficient and the braking distance provides the speed (Figure 2) that can directly be used as an input for Figure 1 to give the throwing distance on the same scaled sketch. It is then very easy to go back from the point of rest to the POI just by applying the throwing distance. Speed [km/h] 100 Figure 1 : Relation of pedestrian throwing distance and vehicle speed [1]. 90 80 70 60 50 40 30 20 10 0 10 m/s² 9 m/s² 0,0 10,0 20,0 30,0 40,0 50,0 60,0 70,0 80,0 90,0 100,0 Braking distance [m] Figure 2 : Relation of braking distance and speed depending on deceleration coefficient. Curves of Figure 2 are obtained by plotting equation ( 1) for different values of speed V and braking coefficient a. V² ( 1 ) d = 2a Best fit for Figure 1 leads to the following equation ( 2) V = 3.33 d + x where x is the distance between the positions of rest of the pedestrian and the car (m), d is the distance between the position of rest of the car and the POI (m), V is the collision speed of the car (m/s), and (x+d) is the throwing distance. ROUGH CALCULATION The goal of this step is to find an approximation of the speeds of each vehicle at each phase. They are of course the impact speeds but also the initial speeds (or driving speeds) at the origin of the conflict zone. The estimation of these speeds demands to start from the position of rest of the vehicles and to estimate different parameters for the post-crash, crash and pre-crash phases, in a reverse chronological order. Estimation of parameters for the post-crash phase This phase describes the kinematics of a vehicle just after the collision. In this step it is the speed immediately at the end of the crash that is estimated. Initial parameters are given from the position of rest of the vehicle. The speed is roughly estimated using constant deceleration of the 8 m/s² 7 m/s² 6 m/s² 5 m/s² 4 m/s² 3 m/s² 2 m/s² 1 m/s² 3

vehicle between the point of impact and its point of rest. More accurate estimation can be achieved by dividing the post-crash phase into several sub-phases corresponding either to different deceleration coefficients or to a sudden change in the course direction. All sub-phase parameters are estimated by using the equation of energy conservation. If both members of the equation are divided by the mass of the vehicle, the speed at the beginning of the post-crash phase (or at the end of the crash phase) can be estimated by : ( 3 ) V ² = V e ² + 2ad Where V e : the speed at the end of the phase [m/s] V : the speed at the beginning of the phase [m/s] a : the applied mean deceleration [m/s²] and d : the distance of the deceleration phase [m]. For the last sub-phase, ending at the position of rest of the vehicle, the speed at this point (V e ) is 0 m/s. The distance traveled by the vehicle in the sub-phase can be estimated by a straight line between the position of the vehicle from the beginning to the end of the sub-phase (from and to its center of gravity for example). The main difficulty in this equation is to find a mean deceleration coefficient. In most cases this parameter is not known very well. A solution is to set a range of probable values instead of a single value. Therefore, the speed is estimated by interval, with a maximum and a minimum. In case the vehicle rotates, the equation (3) can be improved by adding a term corresponding to the dissipated energy in rotation. The yaw inertial moment [5] and the angular speed of the vehicle (cf. Appendix) must be introduced in the equation. In that case, the coefficient of deceleration a in equation (3) is given a value between 4 m/s² and 5 m/s². Estimation of parameters for the crash phase For this phase, both vehicles are studied together. The most important parameters to estimate are the speeds at the beginning of the collision. The short time of the collision phase and the low intensity of contact forces between both vehicles lead to neglect external forces. The following equations apply to the collision phase : Conservation of momentum : ( 4 ) m 1 V1 + m 2 V2 = m1v' 1 + m 2 V' 2 Energy conservation : ( 5 ) ½m 1 V 1 ² + ½m 2 V 2 ² = ½ m 1 E 1 ² + ½ m 1 V 1 ² + ½ m 2 E 2 ² + ½ m 2 V 2 ² where m i is the mass of the vehicle i, V i (resp. V i ) is the speed at the beginning (resp. at the end) of the collision phase [m/s], E i is the Energy Equivalent Speed (EES) of the vehicle [m/s], and i designates the vehicle number. Parameter E i is defined as the crash speed of the same vehicle against a rigid barrier producing the same crash energy than the energy dissipated during the collision. E i are unknown values except if a catalog of crash tests is available. The car deformations are compared to the car deformations in the catalog and an EES can be empirically determined. The first two terms of equation (5) represent the total energy before the impact, and other terms represent the energy repartition between the energy lost in the crash by deformation and energy dissipation, and the rest of energy for each vehicle. Conservation of momentum (4) is a vectorial equation. This equation must be projected on the 2 horizontal and vertical axis. If angles corresponding to the speed direction associated are introduced, equation (4) can be split up in the two following equations : ( 6 ) m 1 V 1 Cos(A 1 )+ m 2 V 2 Cos(A 2 ) = m 1 V 1 Cos(A 1 )+ m 2 V 2 Cos(A 2 ) ( 7 ) m 1 V 1 Sin(A 1 )+ m 2 V 2 Sin(A 2 ) = m 1 V 1 Sin(A 1 )+ m 2 V 2 Sin(A 2 ) where A i is the angle [degree] corresponding to the direction of the speed V i (at the beginning of the crash phase) and A i the angle [degree] associated to the direction of the speed V i (at the end of the crash phase). The resolution of a system of 3 equations assumes that there are 3 unknown factors. In some cases (head-on or rear-end collisions) the projection on the perpendicular axis of the equation (4) is not relevant. Therefore, only 2 equations are available, and only the two collision speeds can be estimated. In general, equations (5), (6) and (7) set a system of 3 non linear equations. Solving this system demands to set 3 unknown factors. Two of them are the speeds for each vehicle just before the impact, one additional unknown factor can be selected among the other parameters of the equations, depending on the information available. Other parameters are available by data collection. However, some of them are not accurate. Therefore it is preferable to give them an interval of probable values instead of a single value. Doing so, the resolution of system (5), (6) and (7) demands to utilize a Monte Carlo simulation that solves the system for every subset of values of parameters sampled randomly from their uniform distribution in their interval. Hundreds or even more simulations can be performed to finally estimate unknown parameters by intervals. This method finally allows to estimate an interval of values for crash-in speeds. The results depend on the number of simulation and the amplitude of the interval of parameters used as inputs. Estimation of parameters for the pre-crash phase The equations used for this phase are the same as those used for the post-crash phase. Brake marks evidence allows to go back in time until the beginning of brake marks. By knowing crash-in speeds, by setting a value for deceleration and by using equation (3), speeds at the beginning of the road marks are easily estimated. In most cases, it is assumed that the beginning of the braking marks on the road corresponds to the time when the wheel blocked. Going further back in time requires specific expert assumptions. Time of braking system activation and reaction time related to the driver can be added. During time of braking system activation (estimated between 0.2s and 0.3s for cars), deceleration follows a linear shape from the initial value to the braking value. A mathematical approximation of the deceleration values during the system activation is, on average, the mean deceleration between the first and the last deceleration values (Figure 3). 4

Longitudinale deceleration (m/s 4,00 2,00 0,00-2,00-4,00-6,00-8,00-10,00 52,00 53,00 54,00 55,00 56,00 57,00 58,00 59,00 60,00 61,00 62,00 experimental curve Figure 3 : Example of deceleration curve The reaction time is often estimated as 1 second. This time takes into account the reaction time itself (0.6s to 0.8s) and the time of feet displacement from the throttle pedal to the braking pedal (0.2s to 0.4s). Reaction time can be increased or reduced depending on the driver (drunk driver, young or old, female or male,...) Adding brake system activation time and time of reaction allows to go back in time a little further than the braking action. It is still possible to go further when analyzing cognitive aspects (visibility, perception, comprehension, etc.) of the accident at its early beginning. SIMULATION OF THE ACCIDENT Rough calculations often issue estimations of the initial speeds. It does not restitute the dynamic behavior of the involved vehicles such as the yaw angular velocity or the pitch angle that can influence the crash condition. These first approximations have to be improved with more realistic three-dimensional equations. These equations are implemented on specialized reconstruction soft wares. Most of them simulate the accident in the course of time and demand inputs such as initial speeds for cars. A few of them can analyze pre-crash phases and especially the vehicles dynamics. We use Pc-Crash [10]. One of its advantages is to simulate the driver s actions such as acceleration, brake, steering wheel or lane change. These are sub-phases that are defined for each vehicle and are put in sequence. They can also be combined (brake with steering wheel action for example). The course of a vehicle is estimated from the initial speed at the beginning of the pre-crash phase and for the consecutive defined sub-phases. For each subphase some parameters must be estimated. These estimations are mean values applied for a sub-phase even though all parameters can take different real values throughout a sub-phase. Two kinds of simulations are possible. The first one consists in starting the simulation from the crash phase and calculating the post-crash motion. This type of simulation is done when no dynamical characteristic of the car i.e. yaw or roll angular velocities plays a role in the pre-crash process (at junction for example). The vehicles are then put in the collision configuration, overlapping one another. The contact zone corresponds to the maximum deformation for each vehicle. The input parameters are the impact speeds obtained from rough calculation, the location and orientation of both vehicles. For the collision calculation, the location of the POI, and the direction of the contact forces have to be defined. The POI must be put somewhere in the overlapping area and defined from EES Time (s) estimation curve approximation curve values or from the total length of the maximum deformation. To describe the post-crash phase, one sequence (or more) is (are) defined to take into account the deceleration of the vehicle after the impact. The simulation is done from the crash phase to the position of rest of each vehicle. The challenge is to estimate a set of parameters in the sequence in order to make coincide the simulated final vehicle positions with the observed position of rest. Once done, the pre-crash phase is simulated. The simulation is performed going back in time to obtain speeds and the location of the vehicles at the beginning of the accident situation. The second method consist in simulating the accident from the beginning of the pre-crash phase to the position of rest of the vehicles. This type of simulation is done for accidents where the vehicle dynamics play a role (loss of control, high level of lateral acceleration, etc.). The main step is to define the initial vehicle position where the stability of the vehicle can be assumed (no yaw, no pitch, no roll), i.e. before the vehicle enters in the accident situation. The exercise is to determine sub-phases and associated parameters to provide a good kinematic simulation. This estimation is realized under constraints such as marks left by tires on the road (the vehicle s path must be in accordance to the marks!), interviews (for driver s actions), and speeds estimated in the rough calculation step. A CASE STUDY Notation: The crash phase is considered as the phase 0. Sub-phases in pre-crash phase are noted in negative order while sub-phases in post-crash phase are described in positive order. Sub-phases 2 1 0-1 -2-3 Phases Post-crash crash Pre-crash Story : The accident involving a Peugeot 205 happened in the morning on a dry road. As the female driver was at the end of a left curve (mean radius 450m), she looked for something in her bag on the passenger seat, and she did not look at the road. Doing so, the car went on the right edge of the road. The driver gave an abrupt steering wheel action to the left to return on the carriage way. The car came back on the road but was going on the opposite lane. To correct her trajectory, she steered to the right and then lost the control of the vehicle. The car left the road on the right edge and the front of the car collided with an embankment. The car did a rollover for 17m (figure 4). Figure 4 : Sketch of the accident The initial speed is estimated by the driver around 90-100 17 m 20 m 30 m 32 m km/h. There were 3 occupants in the car. The driver (without her seatbelt) suffered minor injuries, and two boys, one in front passenger seat with his seatbelt was not injured, whereas the other one in a rear seat without his seat belt suffered minor injuries. For rough calculations, the accident is divided into 2 phases : the crash and the pre-crash. In rollover cases the post-crash and crash phase are identical. To estimate the 5

speed at the beginning of the rollover, equation (3) is applied with a deceleration coefficient 7 between 4 m/s² and 5 m/s², and for a distance from the beginning of the rollover to the position of rest (17m). Starting from the crash phase and going back in time, the pre-crash phase is divided into 4 sub-phases : Sub-phase -1] slid marks on the right edge (20m) Sub-phase -2] skid mark on the road (30m) Sub-phase -3] skid marks on the right edge (14m) Sub-phase -4] roll mark on the right edge (18m) For those 4 sub-phases equation (3) is used with deceleration coefficients estimated by considering the prints of the marks on the site and driver testimonies. For instance the 1 m/s² to 2 m/s² deceleration corresponds to a throttle off (sub-phase 3). From rough calculations, the initial speed of the car is estimated between 80 and 90 km/h. sequences deceleration (m/s²) distance (m) speed (m/s) Min. Max Min Max Average crash 4 5 17 12 13 12,5-1 5 6 20 18 20 19,0-2 2 3 30 21 24 22,5-3 1 2 14 22 25 23,5-4 0 0 18 22 25 23,5 Table 2 : Rough calculation results RESULTS AND DISCUSSION The simulation with Pc-Crash begins at the end of the curve, a few meters before swerving to the right edge. The retained initial speed is 90 km/h. The challenge is to define adequate sub-phases corresponding to the driver s actions, and to improve estimation of kinematics parameters (deceleration, wheel angles) by making the car go through its slid and skid marks. First a sketch is drawn with all important infrastructure elements (road measurements, edge, ditch, curve, marks, position of rest). This sketch is imported in Pc-Crash and adjusted at the good scale. Then, a vehicle Peugeot 205 is selected in the cars database. Vehicle parameters are set to have a vehicle with the real world characteristics (weight, geometry, etc.). The vehicle is put on the sketch at the end of the curve, a few meters before the mark on the edge, with an initial speed of 90 km/h. A small angle (5 ) is given to the vehicle to introduce swerving to the right edge. The directions of the vehicle and the speed are the identical (no yaw angle). The challenge is to find the more realistic simulation for this accident. The pre-crash phase is divided into 5 sub-phases corresponding to the driver s actions: Sub-phase -1] Steering to the left Sub-phase -2] Steering to the right Sub-phase -3] Steering to the left Sub-phase -4] Swerving to the right edge Sub-phase -5] Driving The main difficulty is to estimate the duration of each subphase and the steering angles and velocities. Series of simulation are performed by giving different values to the main parameters (steering angles, deceleration) in order to make the vehicle goes through the tire marks that it left on the road and in the edge. The values are often compared to values obtained in similar experimental conditions (simulators or on track facilities) in order to fit the situation as close as possible to what surely happened. The best solution is obtained when the car follows the course suggested by the marks and the driver s actions seem to correspond to the driver s testimony as far as the psychologist states that the testimony is close to reality. Table 3 : Pre-crash table built from Pc-crash simulation Quantitative Parameters Behavioural Parameters Sub-phase Time Accel. Speed Distance Transv. Control Visibility Interpretation Action (s) (m/s²) (km/h) To POI (m) Position (m) (0/1) (0/1) (S/R/D) Impact 0.0 47 0-5.2 0 1 Danger - 1-1.3-3.3 63-19.5 0.2 1 1 Danger Braking and Steering wheel at left - 2-2.9-1.7 74-51.1-1.8 1 1 Risk Steering wheel at right - 3-3.9-1.2 78-72 -1.2 1 1 Risk Steering wheel at left - 4-4.7-1.4 82-90 -0.48 1 1 Safety Search in her bag an object - 5-5.2 0 85-102 0 1 1 Safety Drive This accident is typical of steering/counter steering actions in curves (20 % of accidents in curves). The driver steers left, steers harder left again 1 and then right whereas her actions are not well-adapted to the situation. The car comes back to the road and recovers its adherence on the four wheels. The driver maneuver (over steering) becomes inadequate and has to be corrected to make the car drive back to its lane. Table 4 : Steering action estimated simulation Sub-phase Angle ( ) Steering wheel action Angular velocity ( /s) - 3 41 137-2 102 185-1 115 230 The magnitude of this third action (steering right in phase 1) is generally larger than the former ones (higher angle velocity and higher wheel angle, see table 4). In most cases, these two actions (over steering left and steering right) leads to a loss of control [2]. According to a preliminary analysis of the accident, the precipitating event is a task other than driving (look for something in the bag on the passenger seat) that made the car leave its lane, followed by a wrong reaction by the driver. The 4 sub-phases of the pre-crash phase are illustrated by one screen view extracted from the reconstruction animation (Figure 6). Car velocity, steering wheel angle as well as the transversal car position (defined by the center of gravity CG) to the middle of the lane are given simultaneously (figure 5). It highlights that 1s. before the crash (within phase -1), the car velocity was 60 km/h, the steering wheel angle 0 and car CG located at 0m. At that moment the dynamic car situation (figure 7) is also given by the longitudinal acceleration ( 2m/s²), the transversal acceleration (-6.5m/s²) as well as the yaw angular velocity (-5.5 /s). According to car dynamic properties, it can be stated that the car is spinning out and cannot be controlled under these conditions. 1 This is partially due to the lack of reaction of the vehicle to the driver s first action, essentially caused by an adherence gap between the pavement and the grassy edge. The driver feels like she has to amplify her action for the car goes the direction she would like it to go 6

Crash avoidance investigation studies must focus on these approaches in order to identify the best automatic car action (braking and/or steering) able to be done when just getting into a phase of car uncontrollability. The goal is to restore the ability of the car to react properly, either under driver actions or automatic ones if necessary. S : Speed (km/h) alpha_sw : Steering w heel Angle (deg) 100,0 80,0 60,0 20,0 0,0-20,0-40,0-60,0-80,0 PHASE -4 Left EDGE 40,0 Opposite Direction Vehicle Direction Right EDGE Right DITCH Right EMBANKEMENT -5,00-4,75-4,50 PHASE -3-4,25-4,00-3,75-3,50-3,25-3,00-2,75-2,50 Figure 5 : Steering wheel angle, Speed and Trajectory of the CG during time on the pre-crash phase. Sub-phase -4 Sub-phase 3 The vehicle swerve on the right edge Time (s) PHASE -2 T (m) S (km/h) Alpha_SW ( PHASE -1-2,25-2,00-1,75-1,50-1,25-1,00-0,75-0,50-0,25 0,00 Steering wheel action on the left Sub-phase -2 Sub-phase -1 Steering wheel action on The driver loose the the right control of the vehicle Figure 6 : Pictures of Pc-Crash animation 9,00 7,00 5,00 3,00 1,00-1,00-3,00-5,00-7,00 T : Transversale position of the du CG from the middle of the lane (m) Deg/s 20 15 10 5 0-5 -10-15 -20-25 -30-35 Right Left Figure 7 : Results of accelerations and yaw angular velocity curves given by the Pc-Crash simulation CONCLUSION 1 2 3 4 5 Transversal acceleration Longitudinal acceleration Yaw angular velocity Sub-Phase -4 Sub-Phase -3 Sub-Phase -2 Sub-Phase -1 The objectives of in depth accident investigations oriented towards primary safety concerns are mainly the understanding of the accident mechanisms, the detection of typical safety problems and the evaluation of the potential effectiveness of on board safety devices. The investigations demand that specialized and highly experienced accidentologists move immediately to the scene of an accident as soon as possible along with the rescue services and the police. Most of the information needed to try to understand is still there but is gone after the road is given back to traffic and the involved persons are transferred to a hospital and have time to forget or tell another story about the accident. What happened? This simple question asks for a tremendous amount of work to collect and analyze the information available about the involved persons, the road, the environment, the vehicles, and last but not least the interactions between all. The understanding of an accident needs a theoretical analytic model. A simple way to present this model is to divide the accident into 4 phases : driving situation, accident situation or pre-crash phase, collision situation (crash phase) and post collision situation (post-crash phase). These phases can also be divided into sub-phases where specific events occurred. The principle of reconstruction is to allow the accidentologist to estimate relevant parameters that characterize those phases or subphases. These parameters are both cognitive and kinematic for the pre-crash phase and are easily gathered in a precollision table for the pre-crash phase which is the most interesting one as far as primary safety is concerned. Speeds are among these relevant parameters (collision speed, speed after the collision, speed at the beginning of the pre-crash phase, etc ), but they are not the only ones. Accelerations, decelerations, angles, vehicle motions, drivers actions perception and comprehension are some of the other parameters that are interesting to know. Rough calculations often give speeds, starting from the position of rest and assuming values for decelerations and vehicles courses. These speeds are consequently used in specialized software to estimate other parameters. The knowledge of accurate data on pre-crash phases do not only help to understand the accident but also to specify typical accident scenarii to be reproduced in experiments (simulators or experiments in test tracks) or to analyze how and when high technological safety devices could have worked to avoid the accident or to reduce their severity. m/s² 4 3 2 1-1 -2-3 -4-5 -6-7 s 7

A case study is presented after a general overview of reconstruction method used at CEESAR. It obviously shows that techniques much depend on accident configurations (head-on, lateral, rear-end) and that a few assumptions are needed (acceleration, crash angles, EES). Cognitive and kinematics parameters are crossed in order to help the analyst in understanding what happened and how driving situation led to a crash, after particular event or series of events occurred. ACKNOWLEDGMENTS The results presented in this study come from real-world investigations carried out within the important program Véhicule et Sécurité Routière PREDIT) partially funded by the French Ministries. The Laboratory of Accidentology and Biomechanics entrusted the Centre Européen d Etudes de Sécurité et d Analyse des Risques (CEESAR) for its realization. Police road crash reports from the French Gendarmerie and Police constitute a large, good quality documentation source for researchers who analyze and use them for the improvement of road safety. REFERENCES [1] Evans A. K., Smith R. Vehicle speed calculation from pedestrian throw distance. IMechE D02398,Vol 213 part D 441-447, 1999. [2] Thomas C. Hermitte T., Perron T., Le Coz J.Y. Driver Action during Real-World Pre-crash Phases JSAE, Yokohama, May 19-21, 1999 [3] Thomas C., Perron T., Le Coz J.Y., Aguadé V. What Happens on the road before Fatal Car Crashes? 40 th AAAM Conf. Vancouver, October 7-9 1996 [4] Mautuit C., Damville A., Perron T., Le Coz J.Y. Reliability of Driver s statements on Driving simulator Canadian Multidisciplinary Road Safety Conf. 1999. [5] Zeidler F., Schreier H.-H., Stadelmann R. Accident Research and Accident Reconstruction by the EES-Accident Reconstruction Method. SAE paper 850256, 1986. [6] Wood D.P., O Riordain S. Monte Carlo simulation Methods Applied to Accident Reconstruction and Avoidance Analysis. SAE paper 940720, 1994. [7] Orlowski K.F., Bundorf R.T. Rollover Crash Test The Influence of Roof strength on Injury Mechanics SAE paper 851734, 1996. [8] Ishikawa H. Impact Model for Accident Reconstruction Normal and Tangential Restitution Coefficients JSAE paper 930654 [9] Backaitis S.H. Accident Reconstruction Technologies : Pedestrian and Motorcycles in Automotive Collisions SAE PT 35, 1990 [10] Forêt-Bruno J.Y., Trosseille X., Le Coz J.Y., Bendjellal F., Steyer C., Phalempin T., Villeforceix D., Dandres P., Got C. Thoracic injury risk in frontal car crashes with occupant restrained with belt load limiter SAE paper 98S-4, 1998. [11] OECD Road accidents : on-site investigations OCDE, 1988 [12] Ferrandez F. L étude détaillée d accidents orientée vers la sécurité primaire. Presses de l école nationale des Ponts et Chaussées. 1995 [13] Dr. Steffan Datentechnik PC-CRASH Operating Manual Version 6.0 Linz, Austria, 1999 APPENDIX In case of rotation in the post-crash phase the equation (1) leads to : ½mV ² + ½Jw ² = ½mVe² + ½Jw² + mad with w (resp. w) is the angular speed at the end (resp. at the beginning) of the phase [ /s], and J the yaw inertial moment [kgm²]. The yaw inertial moment can be estimated for each vehicle by [5] J = 0.1269 m R L Where m is the mass of the vehicle [kg], R the wheel base [m] and L the total length of the vehicle [m]. If the starting point of the phase is the position of rest of the vehicle, the angular velocity w is 0 /s. The angular velocity at the beginning of the phase (w) can be estimated by : w = mgm r R F sign(f) where g the gravity acceleration (9,81 m/s²), M r the coefficient of friction used for the rotation, and F the rotation angle [degree]. This angle is evaluated between the beginning and the end of the associated phase (Figure 8). + 90 position just after the crash rest position (+) +F Figure 8 : Example of post-crash phase with rotation (-) 0 8