Rush Hour: Beat the House with Car Counting?

Background

I work downtown and hate driving in rush hour traffic. Normally I take the train so I can read or watch movies. But sometimes I just have to drive. Back roads are not an option, and no matter which direction I go there will be gridlock during some part of the commute. It the behavior of other drivers that really drives me crazy. And it's the impatient drivers that often cause the phantom slowdowns (no reference, go do your own googling to verify).

My analytical brain watches traffic on the freeway as one lane speeds up and another slows down, and people drift back and forth trying to game the system and get ahead. My gameplan is to just stay patient and ride it out in the same lane unless there is someone obviously going slower than everyone else. To pass the time I mentally tag various cars to gauge the progress of different lanes and different driving behaviors. My intuition tells me that there is usually little difference between lanes. Traffic incidents are hard to predict and gauge, and there are some areas where changing lanes can result in minor improvements. The left-most lanes tend to run a little faster, but not much, and not always.

During these times I wish my brainpower and visibility could extend beyond mild intuition. What if I had the power to track every car in traffic with near 100% fidelity. And then, what if my brain could run an analysis on the cars and identify shifting traffic patterns. Maybe then I could make informed decisions about when to switch lanes and when to ride it out.

Maybe my brain can't do it, but I bet I could teach a computer to do exactly that.

The Idea

Setup a computing system in my car that can monitor all the other cars in traffic. It would track their speed and location. During rush hour, the system would run an algorithm to identify traffic flow patterns and advise lane changes and speed corrections to improve my rate of travel through the madness.

The most conceptually simple way to tackle this is would be to mount a video camera on the car and send the video feed to the computer. The computer would then analyze the images to identify specific cars and track their positions. For a simplistic example check out this video from the VR Lab at Kyungpook National University in Korea. One drawback from this method is that visibility is limited. You'd need to mount the camera up high to increased it's field of view.

Progress

Just a fun idea for now. And by the time I ever get around to working on it, self-driving cars will be the standard and it won't matter anymore!