INTEGRATION: NAVLAB

Rapidly adaptive vision system benefits human driver

Dean Pomerleau, Todd Jochem, Frank Dellaert and Charles Thorpe


The US Department of Transportation estimates that over 90% of the 3.7 million traffic accidents every year in the US involve some form of human error. For example, at night or in bad weather conditions it is often difficult for a human driver to perceive important aspects of the roadway environment, such as the geometry of the road ahead, the condition of the pavement, and the locations of obstacles.  Even under good driving conditions, drowsiness or inattention on the part of the driver can cause perceptual mistakes which can lead to a crash.  Our goal is to apply the combination of smart optics, computation sensors and intelligent algorithms to support people in the task of driving. Potential examples of such support systems include:


  • A vision enhancement system based on smart optics which gives the  driver a better view of the road ahead at night and in bad weather;

  • A lane departure warning system that monitors the vehicle's position in its lane and warns the driver if he begins to drift off the road due to  drowsiness or inattention;

  • A blind spot monitoring system that warns the driver if he starts to perform a lane change when there is another vehicle in the adjacent lane.


This effort leverages off the Carnegie Mellon Robotics Institute's
experience in the development of intelligent vehicle systems, as part of

Integration: