Over the last decade huge steps towards autonomous driving have been made. Nowadays, advanced driver assistant systems like lane departure warning systems and adaptive cruise control can even be found in mid class cars. One may expect fully automatic driving on highways to reach the market by the end of the decade.
However, the availability of cars with highly automated driving functions in urban and suburban areas is still going to require considerably more time. The lack of a clear structure like lane markers on highways substantially complicates the task of identifying the drivable ego lane. Urban areas are characterized by hard-to-model environments containing a high level of clutter while oftentimes missing reliably and easily detectable lane markers.
In a first step, the goal of this project is to identify the features using only low cost sensors, like cameras and radars, and high level digital maps of the environment, as it is common in GPS systems today. In a second step methods are developed to estimate the lanes based on the previous extracted features.