Xu Baoqiang, director of Baidu autonomous driving product Operation Department, fully shared baidu’s development history in autonomous driving and explained baidu’s autonomous driving technology in detail at the Baidu Open Day event. Xu also shared Apollo Moon, baidu’s fifth-generation self-driving car. Apollo Moon is focused on operational scenarios, and the cost has been reduced to 480,000 yuan, which is only one-third of the cost of common driverless cars in the industry, making it possible for large-scale commercial application of driverless cars.
Talking about the future, Xu said he plans to expand driverless car sharing to more than 30 cities within three years, deploying more than 3,000 vehicles to serve more than 3 million users.
How to ensure the safety of autonomous driving
1，40 thousand kilometers of road test every day, let AI become an old driver
Since the 1950s and 1960s, the concept of artificial intelligence began to rise. With the development of industries, artificial intelligence is deeply changing our lives. Like other industries, automobiles are becoming more intelligent with the continuous development of technology. When it comes to intelligent cars, self-driving technology is a must.
Xu said that autonomous driving has been used in public transportation, retail, tourism, logistics and industry, among which urban road sharing is the most technologically difficult and challenging scenario. However, this scene is closely related to people’s lives, which means that this is a scene with the highest commercial potential.
Baidu started research and development related to autonomous driving in 2013 and launched Apollo Go, a self-driving travel service in Changsha in 2019. By 2020, Baidu’s autonomous driving mileage has surpassed 10 million kilometers.
Xu shared that Baidu’s self-driving vehicles can now easily handle merging into traffic, crossing complex intersections, turning around, changing lanes, and overtaking other vehicles.
So how is the “old driver” refined? According to Xu, there are three core elements: first, to build a complete autonomous driving system; second, to conduct thorough road tests for the autonomous driving system; and third, to realize vehicto-road cooperation to help autonomous vehicles cope with various conditions.
Baidu Apollo has so far been tested on more than 12 million kilometers of roads, and that number is growing at a rate of 40,000 kilometers per day. Through such a large number of road tests, Baidu’s self-learning AI drivers can continue to learn and grow. The more they drive, the better their driving ability becomes, and the more easily they can handle a variety of road scenarios.
However, it is far from enough to judge how to drive only by the self-training of AI drivers. Complex road conditions inevitably lead to the existence of blind spots, which can be well solved by vehicle-road cooperation. When an autonomous vehicle passes an intersection, the sensors placed at all directions of the intersection can send various information collected to the vehicle. The autonomous vehicle can better grasp the situation of the intersection by combining these information and perceive various possible dangerous situations in advance.
2，”kill” the safety officer, AI driver also need to get a driver’s license
Many of today’s so-called autonomous vehicles also require a driver to sit behind the wheel and act as a safety officer in case something unexpected happens. So can this security officer “kill”? Xu said that to “kill” the safety officers, it is necessary to build a strong intelligent bicycle system first, and on top of that, redundancy and monitoring systems should be built. The monitoring system can monitor the condition of the vehicle in real time, and in the event of a failure or accident, the redundant system as a “replacement” can replace the faulty part to achieve the same or similar function, allowing the vehicle to slow down or pull over.
In addition, to “kill” the security officers also need to use baidu to build a 5G cloud driving system. When an autonomous vehicle is stuck in a traffic jam or gets stuck, it will be able to remotely operate the vehicle through the 5G cloud to help the vehicle out of the jam.
5G cloud driving operation site
“From the first day of baidu Apollo’s autonomous driving development, safety first has been our core philosophy and value. “Baidu has done a lot of work to ensure the safety of autonomous driving.” Xu baoqiang said. “On the one hand, the system design should be complete to build multiple safety guarantees for autonomous driving. On the other hand, before the vehicle is tested on the road, it has to go through a complete test system, such as simulation test and closed test, to ensure the vehicle can be safely on the road.”
Xu also said that just like humans need a license to drive, autonomous vehicles will have to pass subject 2 in a closed testing field and Subject 3 in an open road before they can get a license. So far, Baidu Apollo has received more than 220 autonomous driving test licenses across the country.
“AI drivers have a longer visual range, faster reaction times, and are less likely to experience fatigue and negative emotions such as road rage.” “We now have a 99.99 percent success rate of comprehensive delivery on urban roads, which is safer than human drivers,” Xu said.
3，With a cost of only 480,000, Apollo Moon will be oriented to operational scenarios
Baidu has already launched the fourth generation of autonomous vehicles, and this year it launched the fifth generation of autonomous vehicles, Apollo Moon. Apollo Moon has integrated sensors except for the main lidar into the body, giving it a look very similar to a production car. In addition, Apollo Moon’s hardware reliability has been significantly improved, with a five-year warranty that claims 20,000 hours of fail-safe service.
In addition, Apollo Moon’s sensors, including the main lidar, are self-cleaning, enabling smoother driving in difficult weather conditions such as heavy rain and snow. Apollo Moon is focused on operational scenarios, so it also has features that make it easier for passengers, such as an interactive roof screen, automatic door opening and closing, and facial recognition when passengers get on and off the bus, Xu said.
And Apollo Moon has brought the cost down to $480,000, putting it in the price range of a production car on the market. According to Xu, this is only a third of the cost of shared driverless cars in the industry.
Apollo Moon’s four bright spots
Xu baoqiang said there are three main reasons why such low costs can be achieved. One is to rely on its technical progress, to achieve light sensor dependence, the other is to use self-developed computing unit, customized laser radar; Third, cooperate with automobile manufacturers for in-depth customization, and realize pre-loading mass production and mass production.
“With a cost of 480,000 yuan and a five-year warranty period, the cost of each car is only 8,000 yuan per month. These costs are lower than the wages of drivers in first-tier cities and have a huge advantage over what drivers in other cities have to pay to fill up their vehicles. This means that we can move closer to large-scale deployment and commercialization.” Xu baoqiang said.
Since 2019, Baidu has opened Its Apollo Go self-driving ride-sharing service in Beijing, Changsha, Cangzhou and other cities. For example, Xu said, “In Yizhuang, Beijing, Apollo Go has achieved full time and regional coverage, and users can call a shared driverless car to travel at any time through their mobile phones.”
The cost of Baidu’s autonomous driving has fallen, making commercial deployment possible
As we can see, with the development of technology and industry, Baidu has made remarkable achievements in the field of autonomous driving. On the one hand, autonomous driving technology is growing faster and faster, and Baidu’s autonomous vehicles have been able to adapt to a variety of complex road conditions. With the progress of autonomous driving technology and the gradual opening of policies, more areas will be opened up to autonomous vehicles. On the other hand, the cost of autonomous vehicles has also been reduced, and with the improvement of the industrial chain, the cost will be further reduced, which makes it possible for Baidu to carry out large-scale commercial deployment of shared driverless cars. It is believed that sharing driverless cars will soon become common, and by then, each of us will experience the convenience brought by autonomous driving.