Autonomous driving ushered in another major milestone.
At the end of 2022, Beijing approved the first batch of fully unmanned autonomous driving tests, and Baidu’s autonomous driving travel service platform “Carrot Kuaipao” is the first batch of approved objects for this policy.
At the same time, Beijing is also the third city that Baidu’s fully unmanned autonomous driving team has won after Wuhan and Chongqing. The operating hours have been extended to 11:00 p.m. to meet the needs of more citizens traveling at night.
Under the turbulent background of salary cuts, layoffs, and shutdowns in the autonomous driving industry this year, it is not easy for Baidu to deliver such results at the end of the year.
On the other side of the Pacific Ocean, Waymo and Cruise are also increasing their deployment of unmanned operations. Among them, Waymo has opened a 24/7 fully unmanned autonomous driving service to the public in San Francisco and doubled its operating area in Phoenix ( including Sky Harbor International Airport).
It is not difficult to see that the top players represented by Baidu are truly releasing the commercial value and social benefits of autonomous driving through “unmanned”.
Operation blossoms and bears fruit
Compared with last year’s overall popularity, the capital’s enthusiasm for autonomous driving will also decline in 2022. The industry is severely polarized, and negative news frequently emerges—from Argo AI’s shutdown, Aurora’s seeking to sell itself, to Tucson’s future high-level changes, looking back at autonomous driving In this world, magical events have been staged one after another this year, and it is not an exaggeration to say that winter has come.
Under the cold winter of capital, there is a sharp contrast between strong brands of autonomous driving and weak brands, and the Matthew effect of “the strong get stronger and the weak get weaker” is further amplified. Because autonomous driving has never been more than a competition of technology, it is easier for leading companies to gather top talents in the industry, and it is easier to integrate upstream and downstream resources, and it is also considered to be a business model that can quickly run through the ground.
Take Baidu’s self-driving travel service platform “Carrot Run” as an example.
With the launch of Carrot Kuaipao, Baidu unmanned vehicles have been further integrated into public life, and exist in a normalized form of travel similar to online car-hailing, as a supplement to the existing transportation network. In other words, the significance of Carrot Run is not to allow autonomous driving to be superior and far away from the daily travel of ordinary people, but also to respect the basic business logic, so that autonomous driving can have a sustainable business closed loop.
Based on Baidu’s autonomous driving operation practice in the past many years, Carrot Kuaipao will also draw a perfect end in 2022.
This year, Carrot Kuaipao has also made a lot of progress in fully unmanned autonomous driving. It has realized the landing in Beijing, Wuhan, and Chongqing. testing companies.
Just recently, Baidu Carrot Kuaipao was among the first to be approved to carry out fully unmanned autonomous driving tests in Beijing.
Specifically, a total of 10 fully unmanned self-driving vehicles were invested in the Bailuobo Kuaipao, and fully unmanned autonomous driving tests were carried out within the time and route designated by the Beijing Development Zone.
Prior to this, Baidu also announced the large-scale expansion of fully unmanned autonomous driving commercial operations in Wuhan, and achieved major breakthroughs, including expanding the fully unmanned operation area from Junshan New City to the third to fifth ring roads of Wuhan Economic Development Zone , doubled the coverage area; expanded the operating hours to 7:00 am to 11:00 pm to fully meet the needs of citizens for night travel. At present, Carrot Run has an operating area of over 130 square kilometers in Wuhan, covering a population of 1 million.
As of the third quarter of 2022, the order volume of Carrot Kuaipao has exceeded 1.4 million, making it the world’s largest autonomous driving operation service provider.
In the three first-tier cities of Beijing, Shanghai and Guangzhou, the average daily order volume of Carrot Kuaipao is more than 15 times, which is close to the average daily order volume of traditional online car-hailing services. In the Q3 quarter of 2022 alone, Carrot Kuaipao won more than 474,000 orders, a year-on-year increase of 311% and a month-on-month increase of 65%.
It is true that the criteria for judging the maturity of a company’s self-driving technology can be measured from indicators such as road test mileage and takeover rate, and the scale and amount of commercial orders will determine to a large extent whether the company can affect its self-hematopoietic ability and whether it can Get more orders and go further.
Looking back at the end of 2022, Baidu is also very optimistic and confident about its development goals for next year.
On the basis of leading the scale of autonomous driving operations and the trend of unmanned, according to the plan, Baidu Apollo will continue to expand its business scale in 2023, further strengthen its self-hematopoietic capabilities, and plans to gradually increase the deployment of 200 fully unmanned vehicles across the country. Vehicles, aiming to build the world’s largest unmanned operation service area.
Technology is getting ahead of policy. Compared with the international perspective, it is necessary to give innovation the best development environment with greater courage in reform and innovation. Compared with the rest of the world, China can only stay ahead in the new round of international competition for intelligent vehicles and driverless travel services only by further liberalizing its autonomous driving policies.
In 2015, when domestic autonomous driving was on the rise, the country raised the development of intelligent connected vehicles to a national strategic level. After that, the Ministry of Industry and Information Technology, the National Development and Reform Commission, the Ministry of Science and Technology and other central ministries and commissions successively issued relevant policies to boost the development of the industry. Since 2021 At present, domestic policies and regulations that are favorable to autonomous driving are constantly “overweighting”.
Coupled with the support of Beijing’s fully unmanned testing-related policies mentioned above, mature autonomous driving products, such as Baidu’s unmanned fleet, will definitely accelerate their landing in the travel market in the foreseeable future.
Strong technical support
The reason why Baidu is able to take the lead in autonomous driving operations and the trend of unmanned vehicles is largely due to its strong technical support, which is also the most basic bottom line for realizing unmanned vehicles.
Just last month’s Apollo Day, Baidu spent nearly 3 hours revealing its breakthroughs in a number of key autonomous driving technologies, and also comprehensively introduced its perception, predictive decision-making, planning control, and data. Full-chain technical solutions such as closed-loop, maps, and computing power have helped the outside world better understand Baidu’s dedication to technology research and development.
-
Wenxin model
Since Tesla first introduced the Transformer large model to iterate FSD, the large model has become a hot word in the industry, and autonomous driving players such as Baidu and Xiaopeng have increased their layout.
Here we first briefly understand the role that large models can play in autonomous driving iterations: analyze large amounts of data, extract useful information, and provide more accurate perception capabilities; learn historical data, build complex decision-making models, and provide smarter decisions Ability; improve the security and stability of the system by conducting extensive training in a simulated environment.
However, a key problem that players generally encounter at present is how to make the large model truly land and give full play to its maximum potential.
Baidu relies on the Wenxin model to try to solve this problem.
Specifically, the Wenxin large model is divided into an automatic driving perception model and a pre-training model with weak supervision of graphics and text. The former has a parameter scale of more than 1 billion, and the small model is trained through a large model, which makes the generalization ability of automatic driving (perception) stronger; the latter relies on the Wenxin graphic large model to recognize thousands of objects, greatly expanding the semantic recognition of automatic driving Data, such as special vehicle recognition, plastic bags, etc., improve the efficiency of solving long-tail problems of autonomous driving.
Based on the unique advantages of Wenxin’s large model, the iteration of Baidu’s automatic driving system can be further accelerated. At present, Baidu’s self-driving technology generalization ability has improved faster than expected, and the delivery time of the new city technology is only 20 days.
-
Autonomous driving map
The earliest used on the car is the standard map, that is, the SD map, which is mainly used to help the driver navigate. SD maps are mainly composed of coarse-grained road topology information, and the accuracy is generally around 15 meters. However, the requirements for assisted driving and automatic driving on the accuracy of maps and the richness of elements are constantly increasing.
In order to make autonomous driving safer, more comfortable, and more efficient, in addition to richer and higher-precision information such as topology, geometry, and semantics, it is necessary to further rethink and define the maps required for autonomous driving based on industry needs.
Baidu also specially launched the Apollo self-driving map, emphasizing that it has the advantages of a new generation of maps with enhanced knowledge, layered and multi-dimensional, and born for self-driving.
It is understood that on the basis of AI empowering large-scale production, cost reduction and efficiency increase can be realized – the automation rate of Baidu high-precision map construction has reached 96%, which has greatly solved the problem of high application cost; in order to ensure the high safety of autonomous driving , Baidu’s autonomous driving map can also integrate car-side perception data and multi-source maps to generate online maps in real time.
In addition, in order to improve the comfort of autonomous driving, the autonomous driving map also builds a driving knowledge map based on the 12 million kilometers of massive road network and spatio-temporal data accumulated by Baidu Maps and the driving experience of human drivers to help the evolution of autonomous driving vehicles.
In addition to the above two major releases, Baidu also introduced its self-developed AI chip and data closed-loop design concept on Apollo Day, improving the overall intelligence level of the autonomous driving system from multiple dimensions.
In fact, none of Baidu’s technological breakthroughs came out of thin air. Turning back the timeline, Baidu has been in the field of autonomous driving for nearly ten years, and its foresight and determination are far ahead of industry players. Baidu began to deploy autonomous driving in 2013, and launched Apollo, the world’s first open platform for autonomous driving, in 2017.
At present, Baidu Apollo has grown into the world’s most active open platform for autonomous driving, and its autonomous driving road test mileage has exceeded 40 million kilometers.
Of course, the test mileage is only a dimension of measurement, and more importantly, the quality of the data accumulated in the test mileage. Through the collection and utilization of data, Baidu’s autonomous driving technology has rapidly iterated and has been fully recognized – a total of 3,477 patents for intelligent driving have been obtained, ranking first in the world for four consecutive years.
Strengthen the closed loop of commercialization
If the core of competition in the first half of autonomous driving is technology, then the second half is a game around the realization of technology.
In the early days, players could invest a lot of manpower and material resources in research and development of technology, but players also had to face the problem of survival. No matter how good the technology is, if it cannot be realized, it is tantamount to a castle in the sky, and the development of the industry will become unsustainable.
Judging from the current status of autonomous driving, Baidu has entered the stage of accelerated operation of the “technology-product-profit-reproduction” commercial closed loop, that is, how to transform technology into commodities and then achieve profitability.
Baidu has tried multiple ways to unlock the commercial value of autonomous driving technology. For example, create the RT6, a forward-looking vehicle designed for autonomous driving.
It is said that the RT6 team has been formed since the beginning of 2021, established the Automotive Robotics Department, and invited a large number of automotive industry experts. This is also the key reason why RT6 is different from other modified self-driving vehicles because of its high safety, high quality and low cost.
-
high security
Once it comes to autonomous driving, safety must be the first priority.
In order to cope with various emergencies, most self-driving vehicles in the industry will still be equipped with safety officers in the driver’s seat. But when the industry has fully entered the fully unmanned stage, many system failures cannot be solved by AI algorithms and modified cars alone.
Thanks to the integrated idea of positive design, RT6 can not only make the automatic driving system meet the requirements of functional safety and expected functional safety, but also actively consider the redundant design of braking, steering, power supply, sensors, etc. It has reached the fault diagnosis and risk reduction system integrating the vehicle and automatic driving system.
-
high quality
Different from the previous Baidu Apollo self-driving cars, the RT6’s three-electric core software technology, interior, and self-driving system, from manufacturing plans to quality standards, are all designed and quality-controlled by Baidu itself. This means that the stability, reliability and consistency of the self-driving fleet can be further guaranteed, and it is also convenient for subsequent large-scale test operations.
Moreover, after the iterations of previous generations of models, from the selection of parts to the design integration and manufacturing process, the mass production level of Baidu’s self-driving vehicles is rapidly improving. In addition, the pre-installation stage focuses on AI algorithms and vehicle optimization Deep integration makes RT6 more humanized and comfortable. This is also difficult for other modified vehicles to surpass.
-
low cost
In the travel service market, labor costs account for a very high proportion of business operations. The emergence of autonomous driving, on the one hand, can liberate human labor, and on the other hand, it can also reduce labor costs and bring huge profit margins. But to bring self-driving vehicles to market on a large scale, there will be another difficulty-the cost of a self-driving passenger car may be two or three times that of a normal passenger car.
Compared with the previous generation model, the cost of RT6 is cut in half, only 250,000 yuan, which is equivalent to the price of an ordinary new energy vehicle.
You know, the price of Tesla’s Model 3 in 2022 is also more than 200,000 yuan, and it will cost tens of thousands of dollars more to configure its FSD system. The sharp drop in cost has allowed Baidu Apollo to deploy tens of thousands of unmanned vehicles across the country. In the future, it will be half cheaper to take an unmanned car than it is now.
In addition, there is a huge breakthrough in RT6, the cockpit design supports two modes switching with/without steering wheel. Tesla, Cruise, and Waymo all have plans to produce fully self-driving vehicles without steering wheels or brake pedals. But unlike their vague production and launch schedule, Baidu made it clear that RT6 will be put into trial operation in small batches in the second half of 2023, which is an obvious step forward.
With the advent of RT6, Baidu has also become the first company in the world to mass-produce unmanned vehicles in terms of positive design and independent research and development, running at the forefront of the world.
No matter from which point of view, Baidu has made enough preparations for the normal operation and commercialization of fully unmanned autonomous driving.
Summarize
Autonomous driving will bring new imaginations to transportation and even society as a whole. However, it is not a simple matter to truly release the commercial value and social benefits of autonomous driving.
In the protracted battle for the landing of autonomous driving, the technical feasibility of players can pass the demo verification to a certain extent, but returning to the origin of practical travel tools, if they cannot provide services to a wider public, it will subvert the traditional online booking In the car travel market, autonomous driving will lose its own meaning.
Only when autonomous driving is recognized by the general public in the large-scale and commercial operation stage can it reach the peak of the industry.
Take Baidu as an example. As of the third quarter of 2022, the total order volume of Carrot Kuaipao has exceeded 1.4 million, becoming the world’s largest autonomous driving operation service provider; among them, the Q3 order volume has increased by 311% year-on-year, showing a rapid momentum.
In the three first-tier cities of Beijing, Shanghai and Guangzhou, the average daily order volume of Carrot Kuaipao is more than 15 times, which is close to the average daily order volume of traditional online car-hailing services. At the same time, Baidu is still increasing investment in operating vehicles to meet the needs of more citizens for self-driving travel, and the order volume next year will further increase.
These data all explain the significance of Baidu’s creation of Carrot Kuaipao, an autonomous driving version of online car-hailing, that is, on the one hand, it makes autonomous driving within reach, close to the daily travel of ordinary people, and magnifies the social benefits of autonomous driving; Respect the basic business logic, so that Baidu autonomous driving has a sustainable business closed loop.
Whether it is at the level of commercial operation of autonomous driving or at the level of unmanned trends, Baidu has taken the lead in setting an example for the industry.
Under the joint promotion of Baidu, other autonomous driving players, and car companies, the era of driverless driving in the true sense is accelerating.
Leifeng.com Leifeng.com
This article is transferred from: https://www.leiphone.com/category/transportation/j2oo8uVEwHMcpwWX.html
This site is only for collection, and the copyright belongs to the original author.