When talking about market competition, Su Kuifeng said that Tencent's mentality is still relatively relaxed. He explained that the automobile industry has a long industrial chain, and it is difficult for car companies to lock in a supplier, which is not the same as the traditional one-size-fits-all model of the Internet. Moreover, each player has a differentiated positioning. As long as there is enough competitiveness on the track where he is positioned, there must be market space.
"Su Kuifeng, General Manager of Tencent Autopilot"
Tencent's autonomous driving strategy is consistent with Tencent's grand strategy, which is to do a good job in connection and tool chain. On the basis of this positioning, Tencent is making three major infrastructures: autonomous driving development cloud, simulation platform and high-precision map. With these tools, Tencent hopes to form a business closed loop between the cloud and the car, accelerate the data flow and quickly iterate the algorithm.
The advantage of Tencent's autonomous driving lies in the support of Tencent's organizational structure. Su Kuifeng introduced that at present, 90% of Tencent's autonomous driving needs can be solved in the smart travel department, supported by Tencent Cloud and with self-control ability.
Tencent hopes to cooperate with OEMs and hardware Tier in the future? 1 suppliers form an iron triangle to promote the industry to run faster. Tencent's role in it is as a server, helping customers accelerate iteration and landing software and hardware products.
Tencent Autopilot Cloud Platform System Architecture
The development, verification, testing and evaluation of autonomous driving require huge time and economic costs, and have strong uncertainties and security risks.
Su Kuifeng introduced that Tencent began to deploy autonomous driving from 20 16, and took virtual simulation technology as one of its key businesses. Combined with localized traffic scenes and application requirements, it can provide exclusive digital tools for autonomous driving R&D and testing. Supported by the computing power of Tencent Cloud, the scene evaluation is accelerated by cloud simulation technology, thus improving the efficiency of R&D and testing.
Tencent Autopilot has applied for more than 40 patents, achieving innovative breakthroughs in high concurrent operation and real effectiveness in the cloud, while making progress in tool chain integrity, scene richness and authenticity.
Tencent Autopilot Virtual Simulation Scene 3D Reconstruction Show
In terms of efficiency, compared with the limited computing power of stand-alone simulation test software, Tencent autopilot cloud simulation technology can support parallel computing of more than 10000 scenes, and greatly reduce the running time of 1000 scenes from 2 days to 4 minutes.
At the same time, Tencent builds a city-level virtual simulation world by using the traffic flow data of virtual and real integration, supporting thousands of self-driving vehicles and hundreds of thousands of traffic flow vehicles to run at the same time, and the vehicles in the virtual city are synchronized in real time, which ensures the effectiveness of the test while being efficient.
In terms of authenticity, Tencent Autopilot has developed models including vehicle dynamics, sensors, traffic flow, maps and scene editing, supporting OpenX series and OSI international simulation standards, and solving the problem that some core models of traditional simulation software rely too much on third parties.
At the same time, Tencent relies on the accumulation of high-precision maps, game technology, AI and other fields to realize high-precision automatic 3D reconstruction of scenes based on artificial intelligence deep learning, with a relative error of less than 3cm. In addition, the consistency between the dynamic model in the system and the field experimental data is over 95%.
Virtual Simulation Test Scene of National Intelligent Network Association (Changsha) Test Site
With the accelerated development of autonomous driving technology, focusing on the research and development of simulation evaluation technology, establishing a perfect evaluation system for smart cars has become industry knowledge, and cloud simulation has become the future trend. In addition to testing tools, cloud simulation will also become a management platform for algorithms, high-precision maps, test data and road mining data, which will play a vital role in ensuring data security. (Text/car home? Xiao Ying)