IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 11, NO. 2, JUNE 2010 463 Argos: An Advanced In-Vehicle Data Recorder on a Massively Sensorized Vehicle for Car Driver Behavior Experimentation Antonio Pérez, M. Isabel García, Manuel Nieto, José L. Pedraza, Santiago Rodríguez, and Juan Zamorano Abstract—A crucial factor in traffic safety is driver behavior. A better understanding of driver actions will help in determining the most common reasons for car accidents. Therefore, research in this field helps to reduce accidents due to driver distraction. This paper presents Argos, which is a complex and powerfully computerized car to help researchers in the study of car driver behavior. The Argos system is an improved in-vehicle data recorder (IVDR) that allows recording many kinds of alphanumerical data such as the speed (vehicle data), the point of gaze (driver data), or the current distance to lateral road marks (environmental data). In addition, Argos can record up to nine simultaneous video images which are synchronized with the alphanumerical data. Argos can also generate and record different kinds of in-car light and audio stim- uli, allowing an experiment supervisor to interact or to schedule specific actions to take place during an experiment. Index Terms—Advanced Driver-Assistance Systems (ADAS), Advanced Vehicle Control and Safety Systems (AVCSS), driver behavior, feedback, in-vehicle data recorder (IVDR). I. I NTRODUCTION F OR THE last 20 years, improving safety in vehicle traffic has been an important objective that has led many insti- tutions and companies (governmental traffic agencies, vehicle manufacturers, etc.) to invest significant amounts of resources, mainly in improving road infrastructure [1], [2] and vehicle computerized subsystems [3], [4], with the purpose of reducing the loss of lives and the financial impact of car crashes. How- ever, another crucial area of research is focused on analyzing driver behavior. In Spain, factors such as distraction, drowsiness, and ex- cessive speed are reported to have the strongest influence on accident statistics. The increasing use of devices such as mobile phones and GPS navigation systems has magnified the problem because they introduce additional sources of distraction [5]. Understanding driver behavior is therefore one of the most Manuscript received March 24, 2009; revised October 14, 2009 and February 1, 2010; accepted March 12, 2010. Date of publication April 19, 2010; date of current version May 25, 2010. The Associate Editor for this paper was A. Broggi. The authors are with the Department of Computer System Architecture and Technology, Technical University of Madrid, Facultad de Informática, Campus de Montegancedo, 28660 Madrid, Spain (e-mail: aperez@fi.upm.es; mgarcia@fi.upm.es; mnieto@fi.upm.es; pedraza@fi.upm.es; srodri@fi.upm.es; jzamora@fi.upm.es). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TITS.2010.2046323 important research issues to improve traffic safety. This area involves research both on psychological aspects (human behav- ior) and automotive embedded system design and development, the latter aiming at providing behavior researchers with power- ful and flexible experimental vehicles. There is a large body of research on characterizing driver behavior, part of which uses data collected from driving sim- ulators. Sekizawa [6] uses these data to model human driving behavior based on a stochastic model. Liang [7] proposes a method to detect driver distraction in real time using eye movements and driving data, mainly the steering wheel angle and lane position, while drivers interact with an auditory in- vehicle system. In the same way, Giusti [8] detects driver sleep-attacks by using the data acquired from the steering wheel movements. Comte [9] focuses his work on attempting to reduce the drivers’ tendency to exceed the speed limits. In his work, targets are presented randomly in a visual scene, providing a further analysis of driver response times to these targets as well as their attention to surprise events. Other studies are based on actual data acquired in specific driving situations such as roadway intersections. Chan’s work [10] focuses on the analysis and synthesis of data acquired with radar sensors from several intersections to improve safety at roadway intersections. Results are verified by means of video images when the radar data are ambiguous or erroneous. Rakha [11] uses an instrumented vehicle equipped with a differential global positioning system (DGPS) and a data- acquisition system to characterize driver behavior based on the perception-reaction time (PRT) at high-speed signalized intersections. The impact of driver age, gender and time to intersection on PRT are also analyzed. The experiment consists in a controlled-road study with no surrounding vehicles. The data acquired are mainly the vehicle speed, the pressing force on the accelerator and brake pedals, and the signal indication coming from a controller box installed at the intersection. Other authors like Ma [12] and Doshi [13] use instrumented vehicles to acquire data to characterize certain aspects of driver behavior. The former builds a simulation model of driver behavior in car-following, and the latter tries to predict intention to change lanes by analyzing the driver’s head movements and eye gaze. Other kinds of systems that use actual driving data are event data recorders (EDRs). They have been used for many years to record information related to vehicle crashes. These systems record certain information from a vehicle over a short period of 1524-9050/$26.00 © 2010 IEEE