In the past 2021, "autonomous driving" continues to become a hot term in the field of technology. Not only have many cities in China launched pilot operations of self-driving taxis, but more and more models with auxiliary driving functions such as lane keeping and automatic follow-up are also increasing, and major Internet and technology giants have thrown themselves into the intelligent automotive industry represented by automatic driving.
One of the most unexpected is that DJI, a well-deserved "brother" in the field of consumer drones, also officially announced its participation in the intelligent car track at the 2021 Shanghai Auto Show, launching its intelligent driving business brand DJI Car.
From consumer-grade drones to intelligent driving, what considerations are behind DJI's decision? With this question, the reporter recently walked into the test base of DJI vehicles located in Longhua District, Shenzhen.
From drones to smart cars
In the first self-driving brand trailer released by DJI, the protagonist is not a car, but a horse. It implies that DJI's ideal state of intelligent driving is like "a horse with enough experience to be able to carry out the instructions issued by the rider according to its own perception and thinking, and make corresponding route planning and decisions."
The reporter learned at the test base that at present, DJI vehicles can carry out intelligent driving modifications for different vehicles according to customer needs, and the specific rules and sensor configurations can also be changed according to customer needs. Of course, the vehicles that have been restructured also need to undergo very detailed calibration and testing work.
At first glance, this set of DJI vehicle-mounted practices is not much different from other intelligent driving vehicles already on the market, so where does DJI's confidence in switching from consumer-grade drones to intelligent driving come from?
In the subsequent exchange, DJI's public relations director Xie Khotan introduced the original intention of DJI's entry into intelligent driving. He said DJI has accumulated a lot of technology in the process of developing quadcopter drones, especially in the field of robotics, or robotics and autonomous systems (RAS). "We found that a series of technologies involved in robotics, such as perception, control, navigation, decision-making, and judgment, can be widely used on various automated mobile platforms, and the world's largest stock of mobile machinery is the automobile."
In this sense, DJI's idea of switching from drones to smart cars is "very natural", and the related technical difficulties are not as great as the outside world imagines. For example, Khotan said, "The UAV should control in six directions, and also consider the coupling of the flight platform and the gimbal, while the car only has four directions, left and right, and the requirements are lower in terms of control difficulty alone." Of course, the ground environment faced by cars is more complex than that of drones in the air, and the requirements for control accuracy and safety are higher. ”
He said that when the industry talks about truly achieving automatic driving, there are "four mountains" that must be solved at the same time: perception, computing power, algorithms, and data. The research and development capabilities accumulated by DJI in the process of making drones, such as image recognition, perception algorithms, navigation and obstacle avoidance functions, are common to the capabilities required by on-board intelligent systems; the large amount of data accumulated in the field of unmanned aerial vehicles is also of great help to intelligent driving technology.
In addition, DJI UAV has always attached importance to the problem of power consumption, and related technologies in turn have also promoted DJI to achieve intelligent driving functions with lower power consumption, so it can not only serve the current new energy vehicles, but also transform traditional oil vehicles, and let intelligent driving run in the case of relatively low power consumption.
The route of global autonomous driving is divergent
With the increasing popularity of autonomous driving, DJI is of course far more than the only one who has devoted himself to this field in recent years. From Internet giants such as Google and Baidu, to soft and hardware companies such as Huawei and DJI, to emerging and traditional car companies such as Tesla, Daimler and Toyota, they have joined this promising market.
At present, global autonomous driving is divided into two distinct factions in the technical route. One of them believes that the perfect autonomous driving experience in the future should be that all vehicles on the road are uniformly controlled by the urban traffic intelligence brain, and all vehicles are automatically driven under its command, and all the technical accumulation of the previous level of L4 will be useless. Therefore, it should be iterative as soon as possible in technology, and achieve L4 level autonomous driving across L2 and L3.
With the current mainstream direction of neural network and machine learning development, if you want to achieve L4-level automatic driving as soon as possible, you first need to accumulate a large amount of data. Therefore, Google, which advocates "one step in place", not only built a huge fleet of vehicles to collect data through actual road tests, but also ran billions of miles of road tests through data center simulations to accumulate road test data as soon as possible.
Tesla's current models are only equipped with L2-level assisted driving systems, but with the help of the huge big data feedback from on-the-road vehicles, it can lay the foundation for Tesla to launch L4-level self-driving technology in the next few years.
Daimler-owned Torc is also deploying its next-generation self-driving truck test fleet in New Mexico and Virginia, and is expected to achieve L4-rated autonomous driving in the coming years.
In China, Baidu hopes to share data with everyone through the "Apollo" open autonomous driving platform to improve autonomous driving capabilities. At the 2022 Consumer Electronics Show (CES), Baidu and Jidu announced that Jidu's first L4-class mass-produced model is expected to be available in 2023, and the concept product may be officially unveiled at the Beijing Auto Show in April this year.
Of course, this technical route is relatively aggressive, due to insufficient data accumulation, related technology is also flawed, and no mass production models have been launched yet.
The other is more practical. At present, the assisted driving technology used in mass production models is mainly used in special scenarios such as high-speed and closed roads, and there is an application bottleneck of "spring breeze on high speed, and urban roads are difficult to walk", so they believe that they should start from L2-level automatic driving in a down-to-earth manner.
In this field, the most typical representative is the Japanese car company. Japanese automakers such as Toyota and Mazda plan to equip major models with autonomous driving functions around 2022, mainly L2 level autonomous driving technology, which will still be mainly controlled by drivers during the driving process. Toyota advocates the principle of "mutual assistance and harmony between people and vehicles", and the core goal of developing this system is to improve driving safety, regardless of whether the vehicle has fully autonomous driving ability.
At present, DJI has also chosen this road.
Xie Khotan said that at present, the mass production models launched by automobile manufacturers have L2-level functions such as lane keeping, automatic follow-up, and emergency braking for obstacles in front of them, but most of the auxiliary driving functions provided by these car companies are hardware integration methods using distributed architecture, which means that if a model with lane-keeping functions wants to increase the automatic follow-up function, it is necessary to re-add a complete set of sensors and controllers to achieve, and the changes are very large. The goal of DJI vehicle is to establish a centralized automobile architecture, using a central domain controller and a set of sensors to achieve intelligent driving functions in multiple different scenarios, and it is also possible to conveniently add or delete certain functions according to the needs of car companies.
He said, "At present, the general feedback from the car companies we have contacted is that this function is urgently needed."
In terms of the development idea of autonomous driving, DJI believes that only by doing a good job of the L2 experience, so that it can adapt to the scene more continuously, as far as possible so that consumers and drivers do not have too much intervention, and at the same time get more support in regulations, policies and infrastructure, and then upgrade to L3, L4. "Making L2 a better experience is the key to our path to higher levels of autonomous driving."
Assisting "overtaking in curves"
In any case, autonomous driving represents a bright future, theoretically, it can avoid people's emotions and fatigue, thereby improving traffic safety. Previous statistics have shown that 40,000 fatal traffic accidents have occurred on the highways of the United States, 90% of which are caused by human misoperation, and mature autonomous driving can greatly reduce the accident rate.
Authoritative experts in the industry said that for China, the development of automatic driving also has a unique advantage, when combined with the current booming new energy vehicles, it may be expected to achieve the effect of "1 + 1 is greater than 2".
The expert said that as "software-defined cars" become a trend, new energy vehicles are jokingly called "computers on four wheels" in the industry, and its high demand for digital technologies such as chips, internet of vehicles, and cloud computing is very high, and the compatibility with automatic driving technology is very high. From this perspective, China's development of autonomous driving technology can take advantage of the "curve overtaking" experience of new energy vehicles, combined with China's advantages in digitalization such as big data, and strive to lead the way.
Relevant experts are also quite optimistic about this, "China in the field of data management and applications and gradually go hand in hand with the United States, the most cutting-edge issues the United States is studying, we are also studying, and there is not much behind the United States, which was unimaginable 10 years ago." We are no worse than Europe or the United States when it comes to the management of the most cutting-edge online data and personal data. ”
It is also to see this possibility of "overtaking in curves", China's governments at all levels from the state to the local level have successively introduced policies to support automatic driving. In 2020, the Ministry of Transport issued the Guiding Opinions of the Ministry of Transport on Promoting the Development and Application of Autonomous Driving Technology for Road Traffic, which set development goals for 2025 and encouraged the development of new formats of autonomous driving. Encourage qualified places to carry out trial operation and commercial operation services such as autonomous vehicle sharing, ferry connection, and intelligent parking.
On April 12, 2021, China's first autonomous bus project was officially put into operation in Yongchuan District, Chongqing, and three L4-level autonomous buses were put on the road to carry passengers, marking a major breakthrough in the commercialization of autonomous driving in China.
But experts also caution that unlike computers that can be restarted after a crash and smartphones that can be patched out of bugs, smart cars have much higher requirements for safety and reliability, because once on the road, it is not only the driver who bears the risk, but also the passing vehicles and pedestrians. In addition, autonomous driving is not only a technical issue, but also involves many issues of laws and regulations. For example, how to divide the responsibility of autonomous vehicles after an accident on the road is still blank.
It is precisely because of this that industry insiders are calling for more experts in technical standards, policy makers and legal experts to join the field of autonomous driving, providing a growth space for this new technological product.
Experts say "autonomous driving pilots" may be a more feasible approach. On November 25, 2021, the first pilot commercialization of autonomous driving travel services in China was officially launched in Beijing, and Baidu and Xiaoma Zhixing became the first batch of enterprises to be allowed to carry out commercialization pilots, investing no more than 100 autonomous vehicles to carry out commercial pilot services within 60 square kilometers of the Beijing Economic Development Zone.
Perhaps only China's development of "bold" and "cautious" development ideas is connected to the future of autonomous driving.