Self-driving car technology that would only add $4000 to price of a car wins Intel science fair

by

Ionut Budisteanu, 19, of Romania was awarded first place for using artificial intelligence to create a viable model for a low-cost, self-driving car at this year’s Intel International Science and Engineering Fair. His whole system should work for no more $4,000.

Ionut created a feasible design for an autonomously controlled car that could detect traffic lanes and curbs, along with the real-time position of the car.

“The most expensive thing from the Google self-driving car is the high resolution 3-D radar, so I was thinking how I could remove it,” he told NBC News.

His solution relies on processing webcam imagery with artificial intelligence technology to pick out the curbs, lane markers, and even soccer balls that roll onto the road. This is coupled with data from a low-resolution 3-D radar that recognizes “big” objects such as other cars, houses, and trees.

All of this information is collected and processed real time by a suite of computers that, in turn, feed into a “supervisor” computer program that calculates the car’s path and drives it down the road.

Budisteanu ran 50 simulations with his system and in 47 of them it performed flawlessly. In three, however, it failed to recognize some people who were 65 to 100 feet (20 to 30 meters) away. He said slightly higher-resolution 3-D radar should do the trick and still keep costs at a fraction of Google’s.

The advantages of self-driving cars are many, noted Budisteanu. More than 2 million people die each year in car wrecks. An additional 50 million people are injured in traffic accidents. 87 percent of the car accidents are only because of human mistakes.

If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks