Driver Situational Awareness
One of the key technology towards situationally adaptive driver assistance systems is understanding and monitoring driver situational awareness (SA). Current collision warning systems simply make warnings based only on the risk of collision and driver’s awareness and intentions are disregarded. As the result, many users think warning and annoying especially they are already aware of the dangers; in the worst case, they turn off the function. Our technology aims to solve the issue by enabling the system to track driver’s awareness of objects in the traffic scene in real-time.
The research consists of two components 1) understanding/modeling of human driver awareness and 2) tracking of user’s awareness based on driver’s gaze behavior. To achieve 1) we collected driver gaze data from professional drivers and modeled the way professional drivers pay attention to the surroundings. Those objects are considered important actors for safe driving. To achieve 2) we made a machine learning system that can predict driver’s situational awareness based on the past gaze trajectory of the driver. By combining 1) and 2) we can identify objects that are important for safe driving, yet the driver is not aware of. We believe by a warning based on this strategy will make the warning more effective and trustworthy.