It is one thing to develop innovative systems for driving assistance, autonomous robots, and other automated systems, but one way or the other, such systems will have to interact with human beings. Sometimes users of such systems are not particularly trained for the manipulation or understanding of such systems (think about how you learned to use the first cruise control system you had in hand in a car). Hence the need to design intuitive and easy-to-use functions, and to verify this with on-field tests and studies (how can a driver, a pilot, an operator interacts with the system, reacts to alerts, what is the cognitive effect of a given system on its operator?)
Naturalistic driving studies, teams coordination studies and other behavioral-related studies often require the acquisition and logging of multiple, high-bandwidth, and heterogeneous sensors data in order to capture as much information as possible about the subject(s) of the study (driver, team members, operator…) and the complete surrounding environment. This can be setup with multiple sensors such as video cameras, CAN bus data, oculometers, radars, physiological sensors, audio, analog & digital inputs, etc.
RTMaps allows setting up such applications in a few clicks, both for the acquisition + datalogging and the playback functions which allows offline reproduction and visualization of the sequence of events and data streams captured by the multiple sensors for post-analysis of the test sessions. Advanced datalogging features are also provided by RTMaps such as pre-triggered recording (which allows recording only interesting moments of long acquisition campaign sessions some moments before the event, and again some moments after in order to fully visualize what situation led to a given incident or just event, and what followed) or distributed recording which allows visualizing a scene recorded from different and distant point of views (from inside the vehicle, from the infrastructure, from a vehicle nearby…)
RTMaps also provides interesting specialized features such as interfaces with third-party analysis software like Excel® or Matlab®, interfaces with simulators (like Oktal SCANeR Studio) in order to study human behavior in reproducible or dangerous situations, and “time marker” components (which provide functions to detect specific situations manually or automatically - with scriptable events -, and seek directly to those in playback mode during post analysis), etc...
HIGHLIGHTS
Supports sensors like : Video cameras, Automotive sensors (CAN & LIN bus, radars, lidars…), Positioning and attitude sensors (GPS, IMUs…), Physiological sensors (GSR, ECG, heart-rate, etc.), Oculometers (Perctech, Facelab, SmartEye, ASL...), And much more…
Real-time synchronized playback
Time markers
Interfaces with simulators
APPLICATION EXAMPLES
Naturalistic studies of drivers behavior in an instrumented car
Esigelec, an engineering school in Rouen, France, developed a car equipped with:
•4 cameras (AXIS IP cameras) •1 front radar (Autocruise) •1 read radar (Autocruise) •vehicle CAN bus (car speed, steering wheel angle, pedals position, gearbox state, lights states, gas consumption, engine RPM, etc.) •1 GPS •several analog environmental sensors (luminosity, temperature, CO & CO2 levels)
Studies in simulator
1. Bypass between reality and simulation The human factors department at Renault regularly runs experiments both in real cars and in driving simulators in order to assess the cognitive load of prototype advanced driving assistance systems. RTMaps provides them a way to use the same software tool almost transparently in both environments, generating the same kind of datasets (same contents and format of datasets files) and so using the same methodology for their different kinds of studies.
RTMaps components have been developed to interface the SCANeR Studio simulator from Oktal. Such components have the capability to retrieve data from the virtual car and its surroundings, which makes it possible to mix in a same experiment virtual signals from the simulator and real ones from sensors like oculometers, physiological sensors, cameras, …
Finally, the methodology also contains an auto-confrontation session where the drivers can visualize the datasets a posteriori and add their own comments about the situation they encountered. Such comments are also recorded and analyzed.
References
Renault -------------------------------------------------------- PSA Peaugeot Citroen -------------------------------------------------------- Eurocopter -------------------------------------------------------- Queensland University of Technology -------------------------------------------------------- Esigelec -------------------------------------------------------- INRETS --------------------------------------------------------
