Interpreting Baidu's return to Android-powered cars dream from Apollo

　　Lu Xun said: "There are no roads in the world, but when more people walk on them, they become roads." A few years ago, artificial intelligence and autonomous driving were still far-fetched dreams; now "AI+" has overshadowed "Internet+", and artificial intelligence has begun to "outperform" humans in more and more fields. Drivers will sooner or later become one of the jobs that will be replaced by AI. Let's follow the automotive electronics editor to learn more about the relevant content.

　　Search companies and artificial intelligence have a certain "genetic" affinity. Even the "AI fever" at home and abroad was first started by Google and Baidu . They have not concealed their ambitions to seize the travel market and the Internet of Vehicles operating system. Waymo, a subsidiary of Google, has built an autonomous driving platform. Baidu's strategic grasp of artificial intelligence is quite clear. Through "intelligent voice technology" and "autonomous driving technology", it has laid out open platforms for "Internet of Things" and "Internet of Vehicles" - DuerOS and Apollo . The former enables the conversation capabilities of diversified intelligent hardware, and the latter allows car manufacturers and developers to overtake in the field of "autonomous driving".

The olive branch of capital began to extend to Baidu,　　which has been in a low ebb in recent years . In the Q2 financial report, Baidu's revenue increased by 14% year-on-year and its net profit reached 4.45 billion yuan, proving that Baidu's advertising model is still the most reassuring profit model for Wall Street investors. When there is no large platform for artificial intelligence, it is imperative to re-evaluate the value of Baidu. After the National Day holiday, Baidu's stock price soared like a "chicken blood", knocking on the 90 billion US dollar mark for the first time, and it was only a few daily limits away from a market value of 100 billion yuan, and it was forced to beat JD.com, which challenged it a few months ago, by more than 30 billion US dollars.

　　1. Baidu's version of "Apollo Project": Apollo is born for openness and empowerment

　　From the algorithm matching and dispatching of online car-hailing platforms, to the data-based operation and deployment of shared bicycles, to the popularity of self-driving new energy vehicles, the travel market has all led to "artificial intelligence". As AI's advantages of not getting tired while driving, not driving under the influence of alcohol, and making more accurate judgments have completely exploded, the Internet of Vehicles dominated by AI algorithms will eventually arrive!

　　With the popularity of navigation software such as Didi Chuxing and Baidu Maps, drivers have actually become accustomed to having a mobile phone screen next to the steering wheel. The greatest potential of mobile Internet can only achieve "assisted driving", while "autonomous driving" that truly relies on machine algorithm control is still in the experimental and primary stages.

　　Tesla is almost synonymous with self-driving cars, but due to traffic regulations in many regions, there must still be a driver behind the wheel. To some extent, Tesla is an "Apple model" that integrates software and hardware. Its biggest constraint is the closed supply chain and technology system, which has led to low-volume production and high-cost Tesla cars gradually becoming luxury goods. This is far from the ambition of universal access to artificial intelligence, but it leaves a broad track for open systems to compete with it.

　　Baidu, which has been exploring the front line of autonomous driving for a long time, only found the feeling of unstoppable momentum this year. Lu Qi took charge of the Intelligent Driving Business Group (IDG) to open up the core technologies and business resources required for fully autonomous driving. In April 2017, the Apollo project was launched. This is an unprecedented open platform in the field of autonomous driving. Its goal is to allow ecological partners and developers to own their own autonomous driving cars through Apollo. According to my superficial understanding, after connecting to the Apollo platform, your private car will become a "Tesla" in seconds.

　　Why is the capital market optimistic about Apollo's autonomous driving solutions open to automakers?

　　(1) Compared with Tesla's "incremental production", Apollo takes the route of activating "existing cars", avoiding duplicate construction by more car manufacturers, with lower costs and broad market prospects.

　　(2) The OS of autonomous vehicles requires deep technology and product support in terms of intelligent interconnection, high-precision map positioning and path planning, environmental perception, vehicle control, human-computer interaction of vehicle systems, and system security. Currently, Apollo has also mastered autonomous driving capabilities in closed venues, and the barriers to entry for competitors are extremely high.

　　(3) The open source platform for autonomous driving cannot be built by one company alone. It needs to connect the industrial chain including vehicle manufacturers, chip and other hardware manufacturers, map companies, software platforms and cloud data. Apollo already has more than 75 domestic and foreign automotive ecosystem partners.

　　2. Apollo 1.5 "plays big", autonomous driving is really getting closer and closer to us!

　　Xunlei founder Cheng Hao divides artificial intelligence startups into "critical applications" and "non-critical applications." The former pursues multiple 9s after 99.9%. If it cannot be achieved, it cannot be commercialized. The reliability gap between 99% and 99.9% is 10 times.

　　In particular, the fault tolerance rate of autonomous driving is extremely low, and the reliability of autonomous driving is closely related to dynamic factors such as traffic flow, road conditions, and pedestrians. This determines that the autonomous driving OS must undergo crazy iterations in deep learning, which adds considerable difficulty to Apollo's efforts to help automakers customize autonomous driving systems for safe driving in complex urban traffic.

　　Just two months after the release of Apollo 1.o, on September 20, Li Zhenyu, general manager of Baidu's Intelligent Driving Business Group, announced the official opening of Apollo 1.5. This version refines the autonomous driving scenarios, forming five core open capabilities of "obstacle perception", "decision planning", "cloud simulation", "high-precision map service", and "end-to-end deep learning", and supports day and night lane autonomous driving.

　　For example, vehicles can accurately identify obstacles when driving at night. If pedestrians or shared bicycles suddenly cross the road, the car equipped with Apollo1.5 can brake automatically instantly, greatly reducing the rate of road safety accidents. Automotive partners can use Apollo1.5 to conduct deep learning of the autonomous driving system in a simulation environment, accelerate the familiarity with road conditions across the country, and plan the most effective driving path and speed based on real-time high-precision maps.

　　It is understood that Apollo has added 65,000 lines of code from 1.0 to 1.5. More than 6,000 developers around the world have recommended Apollo source code, and more than 1,300 partners use Apollo's open source code. This reminds people of the situation when Google opened the source code of the Android 1.0 system in 2008. After obtaining the source code, mobile phone manufacturers quickly eliminated feature phones. The artificial intelligence autonomous driving OS created by Baidu has something in common with Android's open philosophy, which is the key to Baidu's market value leap.

　　Apollo's powerful implementation capabilities and platform strategy make it no longer appropriate to measure Baidu by the previous search market value standards. Baidu's market value has increased from more than 60 billion US dollars to more than 90 billion US dollars in the past two months, which is inseparable from Apollo's rapid progress in the early stages.

　　3. Can Apollo form an autonomous driving ecosystem across the entire industry chain? It’s still full of variables

　　Take Alibaba and Amazon as an example. On December 12, 2015, Alibaba's market value was about 153 billion US dollars, while Amazon's market value was as high as 315.6 billion US dollars. If we think according to traditional business logic, it is impossible to catch up in a short period of time. But on October 10, 2017, Alibaba's market value reached 472.1 billion US dollars, surpassing Amazon's market value of 471.9 billion US dollars. US stock analysts attach great importance to the strategic layout and synergy of listed companies, which is why super apps such as WeChat, Alipay, and iQiyi have not been listed separately. From another perspective, a listed company does not have to worry too much about the ups and downs of market value. "If you bloom, the breeze will come" is the unspeakable secret of US stock market value.

　　According to Baidu's existing advantages in search, maps, and pan-entertainment, the possibility of its market value falling is getting smaller and smaller; in China, Baidu was the first to make efforts in artificial intelligence and has accumulated the most experience. With the platform benefits of DuerOS (home Internet of Things) and Apollo (autonomous driving), Baidu is more and more likely to show its strength in the stage of 100 billion market value. Whether Baidu can play this smart Internet trump card depends on whether Apollo can penetrate the automotive industry chain and become an "enabler" for the transformation of traditional cars to smart cars.

　　At the mass production level, the core of autonomous driving is the algorithm. Whether the algorithm can be mass-produced and exported to developers and ecological partners restricts the scale of the platform. Baidu's dedicated autonomous driving computing platform ACU has three core AI functions: high-precision positioning, environmental perception, and decision-making planning, and accumulates to form a composite series of products, making the vehicle-mounted autonomous driving computing platform technology mass-producible. The update delay and positioning deviation of mobile maps on detailed road conditions have discouraged automobile manufacturers. Baidu's high-precision map relies on perception data processing and cloud data centers to realize the automated production and update of road condition signs, providing a "live guide" for the widespread implementation of autonomous driving.

　　At the capital level, autonomous driving is a huge "moon landing project" (Apollo is Baidu's version of the Apollo program). In recent years, Baidu's investment in artificial intelligence has increased year by year, and R&D costs account for about 15% of total revenue. Baidu Capital has also begun to exert its strength in some key technologies, acquiring and acquiring companies that can be acquired, and holding shares that can be held. The author has made a preliminary statistics: Baidu has fully acquired Kitt.ai (intelligent voice), Raven Technology (intelligent operating system), Zest Finance (financial technology), xPerception (image processing) and other companies to enrich Apollo's technical strength and shorten the R&D cycle. The most scarce assets are still insight into the operating rules of AI technology and commercial talents. A large number of entrepreneurial talents have been recruited, and it is not an exaggeration for the outside world to call Baidu the "Whampoa Military Academy of Artificial Intelligence."

　　In terms of ecosystem building, Apollo does not build self-driving cars, but helps automakers produce cars with self-driving technology. Currently, 17 OEMs including Hyundai, Udacity, and Pioneer, Tier 1 (first-tier manufacturing suppliers) such as King Long Bus, Changan Automobile, Dongfeng Motor, and JAC Motors, start-up technology companies such as Zhixingzhe and Momenta, and Silicon Valley Udacity have joined the Apollo ecosystem. Developers using Apollo open source code is also a process of Apollo's own growing strength.

　　According to the Apollo roadmap, fully autonomous driving will be achieved on highways and ordinary urban roads before December 2020. Apollo 1.5 and subsequent iterations will also continue to iterate and innovate on the autonomous driving open platform.

　　Conclusion:

　　Baidu's market value has soared precisely because it has found a way to unleash its technological advantages through an open AI strategy. Autonomous driving itself is a trillion-dollar market, and the top priority is to help traditional automakers transform and upgrade, empowering them with autonomous driving technology. The massive data and algorithms generated by autonomous driving itself make it one of the AI ​​applications with the highest threshold and broadest prospects. When the Apollo autonomous driving system matures, the DuerOS platform that connects hardware IoT will be able to join forces with Apollo and shine in the field of Internet of Vehicles. The rapid development of autonomous driving technology will quickly reshape people's travel, at least for students who have not yet obtained a driver's license, it is a great boon.

    The above is an introduction to the automotive electronics industry - interpreting Baidu's strong return of the ace Apollo's automotive Android dream. If you want to know more related information, please pay more attention to eeworld. eeworld Electronic Engineering will provide you with more complete, detailed and updated information.