After many decades of seeing self-driving cars in science fiction books or movies, the self-driving car is now a reality. For years, Carnegie Mellon University (CMU) is known for its research on self-driving cars or autonomous robotic cars. In 2007, CMU won the Grand Challenge, a race of several self-driving cars to go from Los Angeles to Las Vegas where the cars must travel over hundred of very sharp left and right turns in the winding mountain passes with different elevation and dangerous tunnels. After that people said: “It is easy to build a self driving car but in a freeway with thousands of other cars, how can you avoid an accident?” It raised the challenge to another level and after few more years of research, CMU has created a new self-driving car that can avoid accident. With advanced computers and high-quality digital cameras and sensors, it possible to build cars that can drive themselves and prove with 100% certainty, that they will be able to avoid accidents.
Self-driving cars are desirable because computer-controlled systems are more accurate than humans. Sophisticated computers allow cars to travel along narrow streets as well as large freeways smoothly without people at the driving wheel. Self-driving cars also have the potential to reduce the loss of human lives in automobile accidents. The problem is people cause accidents because they all react in different ways to a given situation, and with different response times but for computer control cars, it always reacts according to specific instructions and with the same response times Which leads to Carnegie Mellon’s solution: a distributed control system that runs on every car. This system allows the cars to talk to each other via wireless signal. If one car needs to change lane in the freeway, it send signal to other cars nearby then other cars move to make space. If a car in front has to slow down, the other cars that follow know in a matter of milliseconds and they also slow down too. To ensure the concept work safety, the CMU research team starting with two cars on a single lane, and proved with 100% certainty that the cars could not crash. They then slowly increased the complexity of the system by adding more cars, each time using formal verification to ensure 100% safety. Today, the distributed system can control any number of cars moving between any numbers of lanes. I have seen over twenty self driving cars operate at 100 kilometer an hour and change lanes fast, get off freeway quickly, or stop suddenly but everything work well and no accident yet. It all controls by software that talk to each other and that is why many automakers are interested to hire more software engineers for their future cars. One manager told me: “With sophisticated computers built in each car, we need more skilled software people and in the future we may have more software in cars than in current office.”
Even CMU has shown that self driving cars are safe and now it is just a matter of every car manufacturer agreeing to use the same distributed control system so all the cars can “talk” to others. Few weeks ago, a local congressman starts to use CMU self-driving car to travel from the airport to his home about 40 kilometers away to prove that the car is safe enough. The Cadillac SXR was driven along local roads and highways by a computer that uses inputs from radars, laser rangefinders, and infrared cameras. The car reached speeds of 65 mph in the freeway and slow down when travel through several congestion areas without any problem. After that CMU is flooded with many car companies come in to hire its IT graduates.
- Blogs of Prof. John Vu, Carnegie Mellon University