Professor Wang Meiling's team came sixth in an annual robotic-car race held on a 16-kilometre country road in Chifeng, Inner Mongolia, last month. Their chip-controlled, laser-guided Bot just couldn't compete with the winner, the military's heavily armoured Lion, but Wang, an artificial-intelligence researcher at Beijing Institute of Technology (BIT), believes her technology will be used by one of the first robot cars to roll off a mainland production line.
Why didn't you win?
We faced fierce competition. The race, sponsored by the National Natural Science Foundation, has been held for four years - since 2009. More than 10 research institutes sent 13 teams with cars equipped with the best gear and software. Some vehicles, such as the Military Transportation University's Lion and Xian Jiaotong University's Kuafu, had competed in all three previous races and had accumulated lots of experience. Chifeng was our maiden race, and we were prepared for the worst outcome. Soon after we arrived, a Siberian cold front pushed the temperature in Chifeng to near freezing point. We were worried because our civilian-grade sensors were not meant to withstand such a chill. When the race started, the sensors didn't fail us, but our car displeased the judges at the first turn. It turned the steering wheel rapidly and cut the corner like a race car. Some judges gasped and some raised their eyebrows. To them, passenger comfort was more important than speed. In contrast, the military's giant Lion turned smoothly and gently. That was the kind of driving that impressed the judges. I was not surprised that the Lion won first place because its team had much more experience than we did. After the race I told my students that we had done a good job and we could improve our ranking next year with some simple software tweaks to make the ride more comfortable. After all, the Bot was the only first-timer that completed the entire race. I couldn't help screaming when the Bot overtook some crawling competitors at 40km/h, the highest speed allowed.
Traffic laws are widely ignored on the mainland. Can a robot car deal with such chaos?
We have programmed the Bot to deal with some unique road conditions in China. When I drive, the worst nightmare could be a slow, overloaded cargo truck without taillights on a highway at midnight, or a pedestrian suddenly walking in front of my car, or a massive traffic jam in central Beijing in which a slip of the mind could lead to scratches and quarrels. In such cases, a computer can react faster than a human brain, and an automatic brake never gets tired, unlike your right foot. Computers do not doze off. If robotic driving is available on every car, the number of traffic accidents in China will be reduced significantly and driving in rush hour will no longer be a nightmare in big cities. In Chifeng, we encountered many obstacles that are unique to China. For instance, smoke was released from the side of the road to simulate the massive burning of hay that reduces visibility in many Chinese rural areas. The Bot handled them quite well.
What drew you into this field of research?
BIT has many military-related research projects and my team had spent years developing self-guidance systems for unmanned military ground vehicles and aircraft. With the lab becoming more technologically advanced, we want to bring the technology to the civilian sector. The National Natural Science Foundation began funding robotic-car research in 2008 and we won some funding. But even without the money, we would have built the Bot sooner or later, because we have seen China quickly growing from a bicycle kingdom to the world's biggest consumer of automobiles. The next revolution in the car industry will be computer-assisted driving, and we don't want to miss the opportunity to be part of it.
Robotic-car research in China seems decades behind the United States'. Can you catch up?
The research in the US, such as the projects carried out by Google and Stanford University, has been quite impressive. Stanford's Shelley, for instance, has reached nearly 200km/h on racetracks, nearly double the top speed achieved in China. And the most advanced research in the field has been conducted by US military contractors, and they have kept their latest breakthroughs secret. When robotic cars are ready for civilian use, the US government will likely ban the cream of the technology from being exported to China because they know that Chinese are fast learners. The US would fear China's challenge in this field, especially in the military sector. Therefore, Chinese researchers need to work out some crucial technology to rival the US, either on the battlefield or in the global market. One of our biggest disadvantages these days is that our sensors still rely heavily on foreign components. As soon as we can design and produce the sensors, Chinese robotic cars will challenge their US counterparts.
Can battlefield robot cars drive on civilian highways?
Although the military has the most advanced robot-car technology, the challenges of civilian use would be greater than military applications. The main objective of a military robot car is to conduct surveillance or destroy enemy units. It must be covert. If discovered, it must be able to destroy the enemy. If the enemy cannot be stopped, it must activate self-destruction. The objective of a civilian car is to protect human life. It must take care of the driver and passengers, as well as other drivers and pedestrians. Protecting life is harder than destroying it.
Will your hair turn grey before the first Chinese robot cars hit the road?
At present, we are not trying to produce cars smart enough to replace drivers entirely. We are developing technology that will assist people to drive more safely and comfortably. For instance, driving could be set to automatic during a traffic jam, which would allow drivers to take a break by reading a newspaper or watching a movie. Lasers, radar and image sensors could help active the brakes when drivers fail to detect an obstacle. Our biggest problem at present is that the computer sensors sometimes have the same limitations as human eyes. For instance, when there is glare on a road sign on a sunny day, the computer might have difficulty identifying the meaning of the sign. But I am optimistic that the first commercial robotic cars will appear on roads in a decade or two.
Wang Meiling spoke to Stephen Chen