Home > News content

Tencent auto-driving layout view: high-precision map, simulation, data cloud...

via:博客园     time:2018/11/3 23:32:57     readed:149

腾讯自动驾驶布局全态观:高精地图、仿真、数据云......

November 1st is a day to remember in the autopilot circle.

Nanjing, the Tencent Global Partner Partners Conference, shouted smart travel, the three core technology platforms of autopilot debut.

Beijing Naboxa, Baidu World Developers Conference, announced a high-profile announcement with the FAW Red Flag to release the L4-class self-driving passenger car mass production plan.

On the same day, the two giants "do things", it is inevitable to take out more than one.

On this day, the first self-driving taxi in China (costing millions) was officially released on the road, behind the auto-driving startup Wenyuan Zhixing.

Compared with Baidu, and even the startup's “speed and passion”, the pace of commercialization of Tencent's automatic driving is more stable. But in the eyes of Tencent, this is not a sleepy. First, with car safety as a top priority, Tencent will automatically drive product safety as a major red line; second, Tencent chose a more stable road and positioned itself to “automatic driving enabler”, which is more focused Automated driving technology provides technical solutions for car companies to complete their test landings, rather than autopilot operators.

Tencent has its own attitude towards commercialization. In the view of Su Kuifeng, general manager of autonomous driving, L3 is essentially different from L4. The latter supports the driver to completely release his hands, which requires more technical challenges: on the one hand, automatic driving needs to have sufficient redundancy, safety and self-handling ability. In terms of vehicles, a certain redundant system is also required. From the perspective of cars, autonomous driving technologies and application scenarios, Su Kuifeng judges that in the next one or two years, autonomous vehicles do not have real commercial capabilities.

Tencent’s logic in autonomous driving is a participant. A month ago, in order to transform the consumer Internet to the industrial Internet, Tencent adjusted its organizational structure, established the cloud and smart industry business group CSIG, and paralleled the autopilot, Tencent car and location services to the CSIG business group. In terms of autonomous driving, Tencent hopes to build autonomous driving with the depth of the service provider role, rather than joining the autopilot track in the form of component component suppliers.

Su Kuifeng said that how to cooperate with the industry in the future, how to transfer the advantages of Tencent to enterprises, how human-computer interaction can better penetrate into the automated driving car, how to integrate Tencent service on the car network interface, is the original intention of Tencent to join the automatic driving.

L3, L4 auto-driving car debut

Established a team of two hundred or two people. Compared with Waymo, Apple, and Baidu, Tencent's autopilot team is slightly “smooth”. Tesla CTO once said: “Startups do big things, it doesn’t matter to the number of people”. Strictly speaking, in addition to Waymo, this statement applies equally to large and small companies in the field of autonomous driving.

The results of the 100-person autopilot laboratory are the best interpretation.

Earlier, the outside world’s perception of the Tencent Autopilot Laboratory, which was established in September 2016, only stayed at the level of automatic driving test licenses. At the Global Partner Partners Conference, Tencent debuted two cars that were modified based on their autonomous driving solutions.

腾讯自动驾驶布局全态观:高精地图、仿真、数据云......

One is an L3-class self-driving car based on the red flag H7. In terms of sensors, it is equipped with a camera, a 4-line laser radar and a millimeter-wave radar to complete the L3 automatic driving function on the expressway, and this L3 auto-driving solution faces Mass production, the product will be realized.

腾讯自动驾驶布局全态观:高精地图、仿真、数据云......

Another L4-class self-driving car based on Geely Borui GE is currently in the R&D test phase, equipped with three laser radars and three cameras for intelligent driving in complex environments, and WeChat one-button car, V2X car Road coordination.

Lei Feng network new wisdom to understand that in addition to the vehicle itself and sensor hardware, Tencent has layout in algorithms, simulation and other aspects. Its autopilot laboratory includes six parts: high-precision map, simulation platform, data platform, fusion positioning, system architecture, and core algorithm.

Located in software and service providers, in the past two years, Tencent Autopilot Lab has accumulated high-precision map data based on its integration in vehicle networking, big data, cloud platform, deep learning, multi-source information fusion perception and positioning, and high-precision map production. With the technical advantages of many aspects, such as “one core, two products and three platforms”, we will focus on the development platform of autopilot product development, testing and evaluation verification. And parts & accessories, providing complete autopilot technology solutions and open platform support capabilities for depots and suppliers.

The brief description of the “Trouble” of Tencent Autopilot Business Center is as above. In summary, Tencent plays the role of “helper” in the field of automatic driving. The high-precision map, simulation system and data cloud service platform are the three treasures of Tencent's automatic driving. The three core technology platforms also made their debut at the Global Partner Partners Conference.

High precision map

Earlier than the autopilot lab, Tencent's high-precision maps began to grow as early as three years ago. At present, it has accumulated more than one million kilometers of road test data, and more than 5 million kinds of labeled sample data. In the first half of 2019, Tencent will complete the production of high-precision map data for high-speed and fast roads nationwide.

In this way, Tencent is taking a similar path to most traditional graphics companies, namely, drawing high-speed, high-speed maps of urban expressways with its own collection vehicles. Lei Feng.com has learned that Tencent currently has 15 auto-driving collection vehicles, and the cost of vehicle acquisition and configuration is several hundred thousand yuan.

Compared with other domestic maps, Tencent has “no fear” in high-precision maps. Its wholly-owned subsidiary, which has national Grade A surveying and mapping qualifications, conducts map mapping and can accurately detect the accuracy and error of high-precision maps on Tencent's self-driving cars and simulation systems. Compared with Gaode, only map mapping, relying on Ali's self-driving car for verification, Tencent is more free.

Su Kuifeng also stressed many times that Tencent's high-precision map must be a map of auto-driving closed-loop simulation test and auto-driving vehicle verification to ensure its accuracy and error.

In the future, Tencent's high-precision map and positioning solution for auto-driving can be provided separately and charged for usage fees. It can also be opened to the partners with the simulation and data cloud service platform in a product packaging mode.

Simulation System TAD Sim

In the field of automatic driving, time is calculated in seconds. From Silicon Valley Technology, a group of people with a cross-professional background in automotive and software gathered in Tencent, and started construction last year, creating a set of key technical support for autonomous driving — — virtual and real simulation system TAD Sim.

Su Kuifeng mentioned that autopilot requires hundreds of millions of real vehicle test data from research and development to landing. In reality, autopilot vehicles and traffic sections for road test are subject to various restrictions, and it is not easy to obtain detailed data. And the autopilot hardware and algorithms are updated frequently and need to be re-tested continuously. Simulation testing is even more important.

Based on high-precision maps, with Tencent's game engine, virtual reality, cloud game technology, integrated industrial-grade vehicle dynamics model and rendering engine, supplemented by 3D reconstruction technology and virtual and real traffic flow, TAD Sim can realize automatic driving perception Closed-loop simulation verification of all modules such as decision making and control.

Lei Feng Network learned that TAD Sim is a cross-platform distributed system for the development and verification of autonomous driving systems, including stand-alone version and cloud platform version, covering road scenes, traffic flow, vehicle sensing and vehicles. Simulation model such as dynamics.

In the simulation system, Tencent simulates the scene with high-precision maps acquired in reality, and constructs a complete 3D environment through UAV and ground lidar scanning imaging. This means that in addition to high-precision maps in the simulation system, external laser point clouds and visual information are also included.

At the end of this year, Tencent has introduced more than half of the country's high-speed highway high-precision maps into its simulation system, and will complete the introduction of national highway data by the end of next year. Based on its game genes, Tencent also migrated the multiplayer high-concurrency technology scene in the game to the simulation system to help autopilots accumulate multiple scenes at low cost. Tencent, a game-born player, has an advantage in the visual effects of the simulation system.

In terms of vehicle dynamics models, TAD Sim provides a simple model and supports third-party models and depot-owned models. In terms of traffic flow, based on high-resolution map real road data to simulate traffic flow, Tencent simulation system can automatically edit the traffic flow and add randomized and manually set traffic flow.

Data cloud service platform

The data cloud service platform is the key to Tencent's competition with other competitors. Based on the underlying data storage, simulation platform, and more autopilot capabilities, Tencent provides an open platform for autopilot cloud ecosystem. In layman's terms, Tencent provides a cloud-based autopilot combination for autonomous driving partners.

High-precision map data and simulation platform is a good example.

Su Kuifeng mentioned that the stand-alone version can't help the autopilot system to complete hundreds of millions of miles of analog road test, and the parallel acceleration of the cloud simulation platform is the key. The high-precision map enters the cloud platform, and the security problem of mapping is involved in the data on the vehicle. It is easier for the vehicle enterprise to obtain the map data in real time.

The first smart travel strategy: four horizontal and two vertical one

Tencent Automated Driving has become an important part of Tencent's travel ecology. Together with Tencent Auto, Tencent Location Services, Tencent Security, Tencent AI and Tencent Cloud, Tencent's “ambition” is Tencent's travel ecology.

At the Global Partner Conference, Tencent launched its smart travel strategy for the first time. Based on artificial intelligence, content ecology, high-precision maps, simulation platforms, social systems, mobile payment and other underlying support, it is committed to providing partners with access to the “digital world”. The interface has now formed a business matrix of “four horizontals, two verticals, one middle station”, and covers the full scene service of autonomous travel, shared travel, and public travel.

  • Four-way refers to four business matrices of wisdom travel, including Tencent Fed, Tencent Autopilot, Tencent Location Service, Tencent Travel Code, which is the basic platform for Tencent to assist in the industry upgrade;

  • Two vertical refers to Tencent's massive content platform and security system;

  • As a middle-office, Tencent AI and Tencent Cloud provide the underlying guarantee for the business matrix, helping car companies to establish their own car network cloud platform and super brain, providing a full range of intelligent services.

Lei Feng network new wisdom to understand that Tencent smart travel is Tencent's overall solution for partner output.

Among them, Tencent Chelian is one of Tencent's most mature products in the field of smart travel. Also at the Global Partner Partners Conference, Tencent released AI in car last year. At present, it has reached strategic cooperation with BMW, GAC, Changan, FAW, Geely, Dongfeng and other car companies, and has landed GAC Chuanqi GS4, Dongfeng Liuqi T5, etc. A variety of production models.

This year, Tencent Motors upgraded and released the TAI car intelligent system, which will provide small programs for eco-cars. These small programs do not need to be installed. When the car owners drive through gas stations, parking lots, tourist attractions, etc., the small program will automatically “wake up”. Users can easily operate refueling, parking and other services through voice interaction, from “people looking for services” to “return to service”.

Tencent Auto Union also launched an open platform for the first time. The first batch of 11 companies took the lead in the open platform and became the first partner of Tencent Fed.

Compared with speed and passion, Tencent seems to pay more attention to perfection. Tencent in October CSIG, will automatically drive, Tencent car, location services to the business group. This also means that for Tencent, car associations and autonomous driving are equally important businesses, and they can also glimpse out the “ambition” of their layout travel ecology.

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments