What is Chinese-style driverless driving? Zhongzhixing has the answer

When high-end gamers buy a new game, they usually start the "extremely difficult" mode directly, and they will be invited to the platinum level immediately. If the game is replaced with autonomous driving, China's road conditions will definitely be in the "extremely difficult" level. Having a solution to solve the Chinese-style autonomous driving also means that it is sufficient to cope with most road conditions in the world.

Just as SAIC, Baidu and other large companies are scrambling to stake their claim, start-ups are joining the fray. Even if the number of unmanned driving mileage in California is higher, it cannot compare to gaining a firm foothold in China. Zhongzhixing, which has just been established for a year, understands this very well. When visiting Zhongzhixing's Nanjing headquarters, the author saw at least five fully armed Hongqi H7s conducting unmanned driving tests in the park. As a company that has only been established for a year, Zhongzhixing's development speed is quite rapid.

Starting Level 4

According to the definition of autonomous driving by SAE International, it can be divided into six levels from Level 0 to Level 5. However, not every level is a progressive relationship between each other. For example, Level 3 autonomous driving can be understood as an improved version of existing autonomous driving, but Level 4 and above are another more complex and complete set of autonomous driving algorithms.

In other words, manufacturers with confidence and technology can choose not to develop Level 3 autonomous driving and start from Level 4.

From SEA J3016

Judging from the Hongqi H7 modified on site by Zhongzhixing, the on-board sensors include at least 4 multi-point laser radars, multiple cameras, millimeter-wave radars, GPS, and inertial measurement units. More importantly, the interior layout is no different from the mass-produced Hongqi H7 interior. The cab does not have any additional auxiliary control equipment, engineering computers, etc., but the wiper gear lever is modified to add the automatic driving function. In addition, an additional red button for emergency stop is added to the center console, which is very common in test vehicles.

The on-site test drive of Zhongzhixing Hongqi H7 was smooth. For passengers like me, there were only three things to do: get on the car, fasten the seat belt, and chat. According to regulations, the on-board engineer still has to sit in the cockpit to deal with possible emergencies. But in fact, the whole process of turning, circling the poles, avoiding obstacles, pedestrians, and parking was done in one go, and the engineer did not intervene at all. Only when the sanitation workers were in the green belt, the vehicle would slow down and the engineer would honk the horn to remind them.

Under certain conditions, highly automated operations are completed, that is, all driving operations are completed by the unmanned driving system, and humans do not necessarily provide all responses. This is a summary of the actual performance of Zhongzhixing Hongqi H7 on site. Does it look familiar? Yes, this is also the definition of Level 4 autonomous driving.

According to Zhongzhixing, the modified Hongqi H7 can achieve a detection distance of 200 meters and can respond to pedestrians, vehicles, obstacles, traffic lights and signs. It can also predict the next behavior trajectory of surrounding objects, equipped with a high-precision map with an error of less than 2 cm. It only takes 0.2 seconds from the sensor detecting pedestrians to making a braking response. It also has data collection, processing, deep learning, and sharing processing solutions with other vehicles. In short, Zhongzhixing can currently achieve long detection distance, behavior prediction, high-precision driving, and rapid response.

If I hadn't been on hand to test drive it, I would never have imagined that this modified Hongqi H7 could be done without any human intervention, from getting on the car, driving, turning, merging lanes, slowing down, to parking.

Target 2021

If we separate the components and discuss the hardware cost separately, the sensor cost of Zhongzhixing Hongqi H7 is not cheap. Referring to the Velodyne lidar adopted by Waymo early on, the price of a single one can reach 75,000 US dollars. However, as time goes by, the prices of autonomous driving parts and sensors, including lidar, are rapidly decreasing. Taking the VelaDome lidar sold by Velodyne at CES2019 this year as an example, the price of a single one has been controlled to 4,000 US dollars. If purchased in large quantities, it is entirely possible to get a 50% discount on the price.

The development curve of LiDAR is similar to Moore's Law in computer chips. Every 18 months, the resolution of LiDAR sensors doubles, while the price drops by half. This means that the cost of autonomous vehicles is steadily decreasing. Therefore, compared with the hardware cost, CIRS is more concerned about the robustness of the algorithm, that is, in most environments, it can provide mature solutions to ensure driving safety during autonomous driving. This also includes extremely harsh environments such as fog, rain, and snow.

For technological breakthroughs, China Intelligent Travel has set itself a time limit of two years, which means that it will be able to complete Level 4 commercial and implementation conditions in 2021, and begin preliminary promotion in shared travel, and then start promoting large-scale applications in 2022.

Backed by China (Nanjing) Zhigu, the construction of the China Intelligent Driving Artificial Intelligence Research Institute (Intelligent Vehicle National Laboratory) has also become one of the forward-looking layouts of China Intelligent Driving. In fact, China (Nanjing) Zhigu has initially formed a relatively complete AI industry cluster. Here are Microsoft Cloud and Mobile Application Incubation Platform, Google Jiangsu Experience Center, JD (Nanjing) Artificial Intelligence Research Institute, as well as the new brand Byton Automobile Global Headquarters and intelligent vehicle R&D and production base. Relying on local government policy incentives and geographical advantages to accelerate the process of local unmanned driving industrialization, China Intelligent Driving has already calculated the direction for itself.

The future is the integration of 5G and AI

According to a survey conducted by Patent Result, a patent analysis company commissioned by Japan's Nikkei News in August last year, there is no Chinese company among the top 50 manufacturers in the world in terms of competitiveness in autonomous driving. However, it is undeniable that China is developing rapidly in the fields of 5G and artificial intelligence, and it is theoretically possible for unmanned driving technology to take advantage of its advantages and overtake others.

In addition to using domestically produced cars as test models, CIRI said it will integrate the upcoming 5G and existing AI technologies into the underlying architecture, and be prepared for the future 5G environment and large-scale artificial intelligence (AI) from the beginning of the design. It even invented a new term, called 5GAI.

Now, CIX has cooperated with China Unicom to develop 5G deployment for autonomous driving. At the same time, it has also started to contact 5G cloud computing platform, and jointly with China FAW and Nanjing University of Aeronautics and Astronautics to promote the integration of industry, academia and research. From autonomous driving to unmanned driving is a complex system engineering in itself, and CIX is very clear that no one can play a leading role in it.

It is no surprise that from chip and solution suppliers to OEMs, the blowout point for Level 3 and even higher-level autonomous driving models has been set between 2021 and 2023. From the perspective of technological advancement, 2 years is not a long time, and it is even a bit tight. Backed by policies, Zhongzhixing is very confident in occupying a favorable position in localized autonomous driving in terms of products and layout. At least, during the autonomous driving test ride, Zhongzhixing showed off its skills through smooth and no human intervention performance.