Learning how to drive in an AI simulator is becoming more common in China
- AI simulators that replicate the driving experience can save operators rental, labour and equipment costs
- To a certain extent, they may replace traditional driving instructors, although students still need to go on real roads at some point
For some students at Chinese driving schools, learning to drive is pretty similar to playing a video game.
At the Guangshen Driving Training Group, one of the biggest driving training schools in the southern Chinese tech hub of Shenzhen, students check in for practice sessions by scanning a QR code.
Sitting in simulators that resemble old-school racing game arcade cabinets, they navigate realistic 3D simulations of a driving test centre environment – complete with roads, traffic lights, pedestrian crossings and even trees and buildings – on large curved screens.
The self-developed AI project, called Lazy Cat, was launched in June last year. Based on data collected from real test cars, the simulation customises training plans based on users’ performance at each session. Users can also simulate driving along different routes and under various weather conditions.
“We are using technologies to transform traditional training methods, save energy and cut costs in fuel, labour and rent,” said Guangshen general manager Li Zhen.
Guangshen is one of several schools in China’s driver training industry – which was worth 140 billion yuan (US$21.7 billion) in 2019, according to the China Highway and Transportation Society – to embrace artificial intelligence recently.
“Driver training used to be a very traditional industry with not much technology involved,” said Guangshen marketing manager Liu Chuhua. “Now, [AI driver training] is basically a countrywide trend.”
Virtual reality teaches Tencent’s self-driving cars
The company, which operates three other traditional driver training schools with over 500 training cars, had opened 8 AI training centres housing nearly 300 driving training simulators near the city’s central business district and major train stations by the end of 2020, and it aims to build 50 more centres by the end of 2021.
Another driving school in Shenzhen, Carshow Driving Training School, has nine AI-powered learning centres in different areas of the city. In September, the company launched an updated version of its simulated cars and robotic driving instructors, or “robocoaches”. So far, over 5,000 people have graduated from the company’s AI classes, according to local state-owned media Nanfang Daily.
Shanghai-listed Eastern Pioneer Driving School launched an intelligent driving training demonstration based in southwestern China’s Yunnan province in July. The company combined intelligent simulators with VR headsets to offer an immersive learning experience for the demonstration.
“Doing basic training … via 4D and VR simulators not only reduces the time for real car training and fuel emissions … but also reduces the operating costs of the entire driving training industry and improves its competitiveness,” Chen Jianyun, the general manager of Eastern Pioneer Driving School, said in a statement on the company’s website.
03:06
China’s self-driving RoboTaxi hits the road
It is no coincidence that many of these futuristic driving centres are in city centres, where there is high population density increasing the potential customer base, but also high rents.
“It is not possible now to open [traditional] training venues in the centre of the city as there is no such big empty space in the city centre,” Li said. “Even if there is, the rent would be extremely high.”
Rent for land in Shenzhen suburbs has increased from about 10 yuan per square metre per month a few years ago to about 20 to 30 yuan per square metre per month now, she said.
According to Li, about 20 training cars can practice in a 3,000 square metre traditional driving centre, whereas Guangshen can fit more than 60 AI driving simulators in several hundred square metres of space, making the hi-tech option much more cost-effective.
Robotic driving instructors are here to make lessons less stressful
Savings in labour and equipment costs are also reasons behind Guangshen’s recent focus on Lazy Cat, Li said. The simulators cost about 30,000 to 40,000 yuan each compared with training cars, which cost around 120,000 yuan, not counting fuel costs. In addition, two or three employees can staff 14 simulators at the same time, whereas a traditional driving school would need to match the number of driving instructors to its vehicles.
While this saves driving centre operators money, the impact on instructors may not be so positive.
“AI-backed software and devices will give users a better experience, and it’s a good marketing tool for driver training companies,” said Sun Mingchun, chief economist of Haitong International Securities Group. “However, this might threaten the livelihood of driving instructors.”
“All driving instructors are very resistant to the AI project because it will replace them,” Li said. “This is the biggest challenge in the industry, but the advancement of technology is something no one can stop, and it’s not just us who are doing this.”
To be sure, aspiring drivers will still have to practice with real cars before taking their driving tests.
Microsoft Flight Simulator isn’t available in China, but it lets you land in Forbidden City
In China, they have to go through four exams before qualifying for a driving licence: two theory tests about traffic rules, basic concepts on vehicles and correct driving habits, one parking and basic driving skills test on a track, and a road test.
Ding Haifeng, one of Lazy Cat’s thousand or so students, told the Post he spent two hours training with the simulator and another 10 hours in a real car before he passed the parking and basic skills test. He is now practising with the AI simulator again in preparation for the road test.
“I feel it’s exactly like a real car. There is no difference, clutch, throttle, foot brake, hand brake, everything is the same,” the 41-year-old renovation worker said. “The only bad thing is that when I stare at the screen for a long time, my eyes will be a little tired.”
Guangshen has ambitions to increase the percentage of its students taking AI-aided classes from fewer than 10 per cent to more than 50 per cent by this year, Li said.
The passing rates for Guangshen students using the AI simulators and those who go through only traditional classes in real cars are currently about the same, Li said. She added that she was confident the passing rates for AI-aided students will surpass those following the traditional route over time as the company upgrades its software and technology.
Would you pay US$1.68 million for a home-racing system room?
“With intelligent simulators plus robocoaches, we want to standardise the learning process,” Li said, adding that human instructors are more likely to have varying standards.
As the technology develops, the impact for traditional driving schools could be “huge”, Sun said.
“Traditional driving training companies have their advantages, but they can be easily replaced by new technologies. If they do not embrace new technologies [like AI], they are likely to be eliminated.”
