Delphine Dubart, Mohamed Kassaagi, Michèle Moessinger, Laboratory of Accidentology, Biomechanics and human behavior
Maëlle Lefort, Gist France
Paper Number 07-0221
Recent cars are more and more equipped with advanced driver assistance systems (ADAS). The design of useful and safe ADAS requires real driving behavior data in particular for their specification and their tune-up. Our study is focused on the improvement of adaptive cruise control (ACC) design. The specification of such a system requires drivers’ profiles using driver’s actions and vehicle dynamic data (speed, acceleration…) as well as information about close traffic in longitudinal regulation situations. An experiment on real road is currently carried out with 120 common subjects driving an instrumented car. To ensure that representative road situations are taken into account, data are recorded in ecological conditions, with common drivers using a non-ACC equipped car on a 250 km real road. Four data types are recorded: drivers’ actions and comments, car dynamic and road environment characteristics. Drivers’ profiles presented in this paper are based on objective data like headways or speed choices in some relevant driving situations. This experimental method has the advantage to allow understanding both the driver’s real need (and not what the technology enables) and his/her real dynamic use of the car. As for any experimental procedure, it is essential to be aware of some biases which could impact the study conclusions. The data collected from this study and also from other ones should enable building an “intelligent” driving algorithm able to classify any driver in a pre-defined category of profile in order to configure automatically the best ACC functioning mode.
Over the past 15 years major technological changes emerged in the field of automotive industry. New advanced driver assistance systems (route planning, obstacle detection, speed control…) equip more and more recent cars.
Most of the time, in the development of some of these systems, only technological capacities are taken into account. Seldom, human factor aspects are gone into detail. The use of these assistances can have adverse effects if the behavior of the driver does not correspond to the one anticipated by designers . In this paper, we focus on the improvement of adaptive cruise control (ACC) design. ACC system uses sensor to detect the presence of a preceding vehicle and to determine its distance and speed. If a preceding vehicle is detected, the speed of the ACCequipped vehicle is adjusted to maintain a preset safe distance or time headway. This kind of systems has not to disturb the driver in his driving task. That is why the specification of such a system requires driver’s actions and vehicle data as well as information about close traffic in longitudinal regulation situations. To build a real world database, our laboratory is conducting a large scale experiment on drivers’ behavior on real road with 120 subjects driving an instrumented car. This experimental method has the advantage to allow understanding both the driver’s real need (and not what the technology enables) and his/her real dynamic use of the car. Collected database helps driving assistance designers to take into account simultaneously what the technology allows and also drivers’ profiles.
Our objective is to constitute a knowledge database of drivers’ real behavior. Only healthy subjects were selected in order to avoid biases due to pathologies. The study includes 120 participants (60 women and 60 men). They were recruited via a local paper and then distributed in three age groups: 20 to 35, 40 to 55 and more than 60 years. Only persons, who drove more than 5000 km/year and had a driving license for more than 2 years, were chosen. As of January 2007, 36 (among 120 foreseen) persons took part in the experiment.
In the study of real drivers’ behavior, two approaches at least are generally used: directly using his/her own car or using one or a few instrumented cars. As the first approach is difficult to carry out and does not permit us to instrument the vehicle as we desire, we have chosen the second method in which all interesting measures can be recorded. Since most people in our sample drive superminis to small family cars, a large family car such as the Renault Laguna (See Figure 1) we used may have interfered with drivers’ habitudes. However it seems mandatory to use a car in the range corresponding to the primary target market of ACC systems, and subjects had a period to get accustomed to driving a bigger car, which should reduce a potential bias.
ACC systems have been designed to be used essentially on motorway or highway. Our 250 km route (See Figure 2) is composed of 80% of these two kinds of road. The first 30 minutes of driving allow the drivers to adapt themselves to the vehicle. For the remaining route, we consider that the driver has a natural behavior. The experiment takes place in daytime during the same hours to limit the bias due to the traffic. We have to take into account that all the subjects did not have the same meteorological conditions.
The experiment takes place on three meetings. During the first meeting, the participants are interviewed by a psychologist and some questionnaires have to be filled before a medical
examination. Driving on a real road is realized during the second meeting. The subjects are accompanied by an experimenter (a psychologist). They drive between 11 a.m. and 6 p.m. including breaks among one of about 2 hours. During the last meeting, another interview is organized. The subjects view parts of the video recordings and have to explain their actions in very specific driving scenarios. Other neuropsychological and personality tests are also realized.
Acquisition of subjective and objective data
For the data acquisition four methods were used in the study: questionnaires, interview, behavioral and dynamic measurements, and video recordings. During the second meeting, the instrumented car was designed in order to measure at a frequency of 100Hz some indicators of the drivers’ actions (use of the brake, accelerator…), car dynamics (speed, acceleration…) and close vehicles thanks to radars used for ACC systems (relative velocity, headway…). A video recording (See Figure 3) of 4 views (visual scene, rear scene, the face and the hands of the driver) encountered along the route was made simultaneously with drivers’ comments. This observation technique, combining a video recording of the driving scene with the simultaneous recording of different indicators, allows an “exhaustive” analysis of drivers’ behavior in all met real driving situations.
To read more : http://www-nrd.nhtsa.dot.gov/pdf/esv/esv20/07-0221-O.pdf