This article is the product of the NetEase Intelligent Studio (public account smartman 163). This piece is the 29th interview in the "AI Hero" series. Stay focused on AI and keep an eye on the next big era!
Author | Ding Guangsheng
I recall that when my parents were younger, they would bring along a thick Chinese atlas from afar and navigate their way to their destinations with constant map adjustments and directional corrections. Later, our newly purchased cars came equipped with stylish electronic navigation systems. However, the outdated and cumbersome interfaces often made navigation challenging.
It wasn't until the arrival of mobile navigation that things changed. Its user-friendly design and timely map updates provided an unprecedented travel experience. No longer did we have to distinguish between northeast and southwest; simply following the "commands" got us to our destination.
However, in the era of autonomous driving, maps are evolving into high-precision maps and high-precision positioning. This will further liberate humanity, allowing us to simply sit back and let self-driving cars take over. You won't even need to worry about maps or giving directions.
So, how will navigation work for self-driving cars?
Gao Dong Map's Executive President, Wei Dong, has a unique perspective. He told NetEase Intelligence, "High-precision maps and high-precision positioning are bound to become essential infrastructure for autonomous driving programs. Sensors, lidars, and other technologies will play an integrative role, but regardless of the approach, we must first focus on building robust infrastructure. Gaode Map has two main directions in the field of autonomous driving: first, leveraging big data and cloud computing capabilities to enhance navigation services; secondly, providing high-precision maps and high-precision positioning to empower both businesses and users."
Talking about building infrastructure for autonomous driving, Wei Dong candidly admitted, "This is a mission-driven effort. Perhaps we won't fully grasp the significance of this work until three or five years from now." As a map service provider, Gaode's strength lies in data accumulation and R&D capabilities, but it lacks the critical high-precision positioning capabilities necessary for infrastructure.
A few days ago, Gaode Map and the positioning company Chihiro Positioning reached a strategic partnership. The latter's core competence lies in offering high-precision positioning services based on the BeiDou satellite positioning system, supported by the construction of the nation's 1450 ground-based enhancement stations. They offer accurate positioning services across most of the country 7×24 hours a day. These two entities complement each other well.
"Building a Car" or "Building a Road"? The Chinese Auto Industry Must Not Repeat Past Mistakes
For the "car builder" or "road builder," old and new players in the automotive industry have their own preferences. Wei Dong pointed out that Gaode Map aims to be the "road builder." The Chinese auto industry has already learned a lesson from the past.
Wei Dong said that now our development of autonomous driving has entered a misunderstanding phase. It's somewhat like when China first entered the automobile era. Back then, China's roads were poor, and infrastructure was underdeveloped. The best-selling cars were Japanese models, especially off-road vehicles. Many areas in today's rural China still sell better cars or Japanese off-road vehicles. Why? The poor infrastructure meant that cars capable of adapting to harsh conditions were favored.
Wei Dong told NetEase Intelligence that this mindset is similar to the ideas of many people working on autonomous driving today. They try every means possible to improve cars with lidar and other sensors. A car that could truly meet high-precision positioning abilities might cost hundreds of thousands of dollars. Gaoding looks toward the future and focuses on infrastructure.
This inevitably relates to the path of autonomous driving. Currently, there are roughly two ways to implement autonomous driving: one relies on machine learning vision schemes, and the other depends on traditional high-precision positioning using high-precision maps. In response, Yang Wei, Vice President of Research and Development at Gaode Map, added, "These represent two schools of thought. One is the self-sensing system, characterized by vehicle sensors detecting the world around them through laser radar and other methods. The onboard control computer acts like the human brain, deciding what actions are needed. Traditional automakers generally follow this route."
The other school is the V2X model, designed for the future of transportation. It enables communication between vehicles, between vehicles and base stations, and between base stations themselves, providing real-time traffic information, road conditions, and pedestrian data.
It is understood that with industry development, there is now a trend of convergence between these two schools.
Secure Autonomous Driving: High-Precision Maps + Positioning + Sensors
Security is an unavoidable topic in the field of transportation. In fact, autonomous driving has caused more and more traffic accidents worldwide. Google's driverless cars have been involved in over a dozen accidents, with a particularly serious incident occurring in January 2016. A Lexus SUV test vehicle collided with a bus.
In addition, last year in January, Tesla's first fatal accident involving autonomous driving occurred in China. A Tesla sedan crashed into a cleaning vehicle on the Beijing-Hong Kong-Macao Expressway, resulting in the driver's death. At the time, the owner had activated Tesla's autonomous driving mode.
Wei Dong said that for autonomous driving, the technology can be categorized into two major parts: the first is perception, and the second is control and decision-making. "The first involves lidar, images, or millimeter-wave devices. It deals with flat surfaces rather than spatial concepts. A car traveling on a road is a spatial concept, so images alone cannot solve spatial problems."
"We recently saw Tesla's autonomous driving accident. While the car was on the highway, the sun was bright, and a white-painted pickup truck was parked on the side of the road. The image misidentified it as a cloud, leading to a direct collision, causing casualties," said Yang Wei.
Regarding the safety of autonomous driving, Wei Dong believes that the implementation of technology is also worth considering. Sensor accuracy is always advancing in a more refined direction. This is an inevitable trend in hardware development. "The problem is that we find that at this stage, the best positioning and navigation model is considered reliable, yet many people opt for a more complex path, such as image recognition and lidar. While these approaches can work, whether they represent the best solution is worth discussion," Wei Dong stated.
Wei Dong further emphasized that sensors have other functions besides positioning and environmental sensing. For instance, they play a role in emergency situations to prevent accidents. Sensors will always be necessary, and the future demands higher precision. The question is whether costs should be considered at this stage? Can the algorithm's efficiency and processing time meet the requirements? Even a few seconds could lead to an accident.
"In the future, high-precision maps + high-precision positioning infrastructure and sensors, lidar, etc., will play complementary roles. The former solves the problem of absolute accuracy, while the latter addresses relative accuracy and safety measures," Wei Dong outlined the path to implementing autonomous driving and ensuring safety.
Next: Achieving Centimeter-Level Positioning - Cars Are Not Passers-by
We've all experienced situations where inaccurate positioning and navigation deviations left us confused about directions. This is frustrating.
As high-precision map positioning becomes increasingly important for self-driving cars, Zhi Lei, Director of Chihiro's Smart Driving Division, said that for autonomous driving, positioning itself is a crucial part of control. Accurate positioning essentially determines the decision-making endpoint. For example, with high-precision maps, a vehicle can determine if it's approaching an uphill, downhill, or a sharp turn. "If the speed is too fast when entering a high-speed corner, the autopilot system will help slow down automatically. This requires the combination of high-precision positioning and high-precision maps. In the long term, autonomous driving will require centimeter-level positioning."
In addition, Wei Dong told NetEase Intelligence that high-precision positioning technology will gradually integrate into Gaode's products, including car versions. Gaode Map's "small goal" is to achieve centimeter-level coverage in China by 2018 and global meter-level services by 2019.
Currently, car brands such as Changan, Geely, Dongfeng Kai Chen, and BYD have pre-installed high-precision car map versions; post-market, Gaode is capturing more market share.
However, it is clear that the high-precision positioning function is currently unavailable on the general public's mobile phones. (End)
Follow NetEase Smart (smartman163) to get the latest reports on the artificial intelligence industry.
Click here to view the "AI Heroes" series, step by step!
1. Interview with Microsoft Hong Xiaowen: In the next decade, AI applications will become a battlefield for soldiers.
2. Interview with Yu Kai, Horizon Robotics Technology: I'm not alone in fearing the industry bubble!
3. Interview with Kewo Si Qian Dongqi: Artificial intelligence landing for a decade, being cool doesn't mean you can survive.
4. Interview with Yun Zhou from Technology Week: A Unique "Fighting Method" for the "National Team" Facing Face-offs.
5. Exclusive interview with U Thant Technology Wu Gansha: Sharing an L5 autonomous vehicle in 2021 is very difficult.
6. Interview with You Jian must choose Zhou Jian: The popularity of humanoid robots will take 10 years, and 90% of entrepreneurs will fail.
7. Interview with Lenovo: You used to be an AI crook. Now I hope AI is only a "slow cow."
8. Interview with the Fourth Paradigm Dai Wenyuan: When will AI's Windows era arrive?
9. Interview with Intel Song Jiqiang: If robots replace your job, learn to be flexible and coexist with AI.
10. Interview with Tucson's Future Hou Xiaodi: Those who can't make profits but claim to want to change the world are either deceiving themselves or lying to others.
11. Interview with Gowild Qiong Nan: I hate free hardware, all payment should be respected, and low-cost robots will be popular this year.
12. Interview with Codelong Technology Huang Dinglong: Artificial intelligence has been cool for more than 60 years now; the core is making money.
13. Interview with Microsoft Huang Xuedong: What is the technological outlook of voice interaction?
14. Interview with Daxie Huang Xiaoqing: When will the robot's 1G era come?
15. Interview with Li Zhifei: Artificial intelligence will become an extension of human intelligence.
16. Interview with ROOBO Liu Yingbo: How do I avoid pitfalls in the robot market?
17. Interview with English fluently said that CEO Wang Hao: Why do Chinese learn English so effectively?
18. Interview with Baidu Ventures Liu Wei: Which AI companies will die in the bubble?
19. Dialogue with Hasabis: The story of AlphaGo's farewell to Go AI has just begun.
20. Interview with Jürgen: AI has long known where the industry will go in the next decade.
21. Interview with Nico: How far is the use of brain-computer interfaces to download consciousness to achieve immortality?
22. Interview with Yang Qiang: Alpha Dog is God? It also has a fatal weakness.
23. Conversation Yuval: AI works irreversibly for humanity and wisdom will control the world.
24, interview Peng Jun: L3 is not suitable for commercial vehicles to consider scene application.
25, dialogue Li Deyi: "Drive the brain" into the car traffic jam will become history.
26. Dialogue View: DoerOS Do Android Play in the Voice Interactive Era?
27. Interview with Xie Dianxia: Don't just stare at the speaker speech interaction will refactor APP.
28. Interview with Jian Renxian: How to Save the Mentally Retarded Robot with Emotional Interactions
AL/CU Universal Terminal Block
Cable Terminal Block,Universal Terminal Block,Waterproof Terminal Block,Low Voltage Terminal Block
ZHEJIANG QIANNA ELECTRIC CO.,LTD , https://www.traner-elec.com