In order to evaluate the impact of driver assistance systems under various situations, researchers have attempted to reproduce accurate traffic situations and accidents by traffic simulations. Here, we propose a new simulator STREET (Safety & Traffic REaltime Evaluation Tool) that has a driver model with a cognition model and a decision-making model in it. This paper mainly describes the aim and the architecture of this novel driver model.
In the cognition model, there are three stages: 1) detecting objects in the field of view, 2) classifying such objects like a lead vehicle or oncoming vehicle, etc., and getting information, and 3) setting the driver’s gaze direction. In the decision-making module, there are two stages: the first stage is to decide a maneuver for each recognized object by using “a decision rule with maps” expressed as the status space region defined by object’s parameters such as distance and velocity as axes. The second stage is to decide the most appropriate maneuver among the combinations permitted in the acceleration/deceleration ranges for each object in succession to the first stage. The driving maneuver is switched in sequence based on the decision-making model output and the vehicle motion is then consequently calculated. When the traffic participants are added in the scene, decision-making rules are added for them, allowing STREET to correspond to complex traffic situations.
Two benefits are expected by using STREET. One is that users can evaluate and understand system activation under the target situations. Another is that the system can be evaluated under various traffic situations beyond the target situations so that the users can assess the limitations of the system. Some preliminary results using STREET and further development plans for the system are also discussed.