U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
The need for cognitive models of the driver was emphasized at the University of Helsinki in Finland, TNO in the Netherlands, and INRETS in France. Present methods and data permit us to know what drivers do, but not why they do it. Employing analytic methods to produce cognitive driver models will help us to develop an understanding of driver behavior. Such models are useful in several ways. They are part of microscopic traffic models that can be validated by observing traffic flow. Indeed, the driver models used for this purpose at INRETS are so detailed that they are referred to as "nanoscopic" driver models. Cognitive models are also useful when implementing human-centered design and analysis. Instead of having to perform a new experiment to answer each new question, the model itself can generate answers.
Figure 28 shows the representation of driver cognitive processes developed at INRETS. This model is written in Smalltalk, a computer language well suited for artificial intelligence applications. INRETS has a considerable financial investment in this model, which has been developed over 10 years with a 3-year break because of other internal priorities. Only now are validation studies being conducted for the model. This delay in validation illustrates how important continuous funding is for high-risk, high-reward basic research. The team congratulates INRETS for seeking and funding such a long-term goal.
Figure 28. The COSMODRIVE cognitive model. (INRETS)
INRETS researchers make an important distinction between behavioral models and cognitive models. A behavioral model focuses on what the driver does. Such models are often descriptive because they can predict behavior but cannot explain it. A cognitive model focuses on the mental activities carried out during driving. It explains why the driver undertakes certain actions. The researchers believe this level of analysis is necessary to understand human errors and difficulties, and to design driving assistance adapted to driver needs.
Many American human factors researchers would not entirely accept this dichotomy because behavior and cognition can be combined in a single model. In such a unified model, the control of the vehicle is called inner-loop control. The control of cognitive activities that guide the strategic reasons for undertaking a trip is called outer-loop control. The FHWA Interactive Highway Safety Design Model (IHSDM) quantitative driver model can combine behavioral and cognitive aspects, although most research to date has focused on inner-loop control.
The COSMODRIVE (Cognitive Simulation Model of the Driver) model uses a computer to simulate human cognitive processes. As a computational model, it draws on a rich history of artificial intelligence models created by a team of computer scientists, psychologists, and engineers. Frames are used as the formalism for representing driver knowledge extracted from experimental results and controlled observation. Each mental process is implemented as a cognitive agent. The greatest strength of COSMODRIVE is its ability to make quantitative predictions. Because the emphasis is on outer-loop control, the model is less concerned with the behavioral mechanics of keeping the vehicle on the road.
COSMODRIVE Contact Information
Figure 29 shows the driver behavior model that guides human factors research at TNO. It draws on the tradition of qualitative information-processing descriptions of behavior and emphasizes control of the vehicle. Research guided by this model stresses behavioral measures such as the following:
Figure 29. Driver behavior model used at TNO.
Driver Behavior Model Contact Information
At the University of Helsinki, Professor Heikki Summala emphasized the importance of using behavioral models to guide roadway safety research. Indeed, in Europe a formal discipline called "traffic psychology" uses such models. Figure 30 shows one such model that relates the level of psychological processing to a functional hierarchy of vehicle control and a functional taxonomy of strategic behavior. These functional divisions are similar to inner- and outer-loop control in quantitative driver models based on the mathematics of control theory. Attention control is a key psychological process involved in driver distraction, overload, and underload.
This model has the advantage of combining the two types of control to make a variety of predictions from high-level trip decisions to low-level vehicle control. It is very helpful in providing a framework that integrates many results about behavioral adaptation, risk taking, maintenance of safety margins, and allocation of attention. Although the model is quite useful, it does not offer the quantitative predictions of a computational model.
Figure 30. The driver task-cube model.
Hierarchical Driver Task-Cube Model Contact Information
Professor Heikki Summala
University of Helsinki Traffic Research Unit
Siltavuorenpenger 20 D
FI-00014 University of Helsinki
Phone: +358 9 19 12 94 20
|<< Previous||Contents||Next >>|