The autonomous driving market is still an arena for giants after all

At the beginning of 2019, Alibaba DAMO Academy announced the "Top Ten Technology Trends of 2019". Looking back a year later, most of the trends still hold true today. There are two related to autonomous driving:

 
Trend 1: Autonomous driving enters a period of calm development.
 
Relying solely on the "single-vehicle intelligence" approach to innovate automobiles will not achieve the ultimate unmanned driving for a long time, but it does not mean that autonomous driving has completely entered a cold winter. The vehicle-road collaborative technology route will accelerate the arrival of unmanned driving. In the next 2-3 years, the commercial application of autonomous driving represented by limited scenarios such as logistics and transportation will usher in new progress. For example, commercial scenarios such as fixed-line buses, unmanned delivery, and micro-circulation in parks will be quickly implemented.
 
Trend 2: Real-time simulation of cities becomes possible, and smart cities are born.
 
The perception data of urban public infrastructure and the real-time pulse data stream of cities will converge on large computing platforms. The development of computing power and algorithms will promote the real-time integration of unstructured information such as video and other structured information. Real-time simulation of cities becomes possible, and local urban intelligence will be upgraded to global intelligence. In the future, more forces will appear to carry out the research and development of urban brain technology and applications. Smart cities with full-time and space perception, full-factor linkage, and full-cycle iteration will be born on physical cities.
 
The development of autonomous driving simulation is closely related to the development of the entire autonomous driving industry. In the past two years, the pace of autonomous driving development has slowed down, and start-ups have faced unprecedented challenges. The same is true for startups in the field of autonomous driving simulation. Sensor simulation company RightHook has not updated its dynamics for two years, and basically no new autonomous driving simulation startups appeared in 2019.
 
On the contrary, the giants are making frequent moves.
 
At the Shanghai Auto Show in April 2019, Huawei's autonomous driving cloud service Octopus was launched, which includes a simulation test platform. In
 
December 2019, Waymo acquired LatentLogic to strengthen its simulation technology. In
 
April 2020, Alibaba Damo Academy released the "hybrid simulation test platform" for autonomous driving.
 
GAC believes that: in the past, the virtual simulation platform played a supplementary role to the real vehicle test platform. At the L3 and above autonomous driving stage, virtual simulation testing has become an indispensable part of the autonomous driving research and development process. At present, GAC's virtual simulation testing accounts for more than 60% of the autonomous driving research and development process, and will increase to 80% in the future. Simulation is
 
not only indispensable in the research and development of single-vehicle intelligence, but also in the research and development of autonomous driving with vehicle-road collaboration.
 
As autonomous driving develops from "single-vehicle intelligence" to "vehicle-road collaboration", autonomous driving simulation has also developed from dynamic simulation, sensor simulation, and road simulation (static) to traffic flow simulation (dynamic) and smart city simulation. 51 VR
 
, which has raised hundreds of millions of yuan , changed its name to 51WORLD after experiencing the VR bubble and began to deploy digital twin cities and autonomous driving simulation. In November 2019, 51WORLD signed a contract to settle in Chongqing Liangjiang New District and will focus on expanding digital twin city innovative application products and autonomous driving vehicle simulation products in Chongqing. In fact, the combination of VR and autonomous driving simulation is not a helpless transformation for 51WORLD. In fact, VR/AR is becoming more and more important in autonomous driving simulation. The technical means of virtual scene construction usually include building scenes based on modeling software, building scenes based on completed games, building scenes based on VR/AR methods, and generating scenes based on high-precision maps. In August 2019, rFpro launched a set of autonomous driving simulation training systems based on VR scenes. This training system has the following characteristics: (1) A series of autonomous driving simulation operations can be completed in this software.
 

 

 

(2) The rFpro simulation driving system supports the import of third-party map models in multiple formats, including IPG ROAD5, .max, .fbx, OpenFlight, Open Scene Graph, and .obj with HIDEF high fidelity and other features.
 
Due to the importance of autonomous driving simulation, the formulation of simulation standards has also been launched.
 
ASAM (German Association for Standardization of Automation and Measuring Systems) is the leader in global autonomous driving scene simulation test standards (mainly OpenX series). Since ASAM launched the OpenX series format standard, more than 100 companies around the world have participated in the formulation of this series of standards, including major vehicle manufacturers and Tier1 in Europe, the United States and Japan.

 
In the field of ASAM simulation verification, the OpenX series of standards mainly include five major sections: Open-DRIVE, OpenSCENARIO, Open SimulationInterface (OSI), Open-LABEL and OpenCRG. OpenDRIVE and OpenSCENARIO unify the different data formats of simulation scenarios; OpenLABEL will provide a unified calibration method for raw data and scenarios; OSI connects autonomous driving functions and simulation tools, and integrates multiple sensors; OpenCRG realizes the interaction between road surface physical information and static road scenes.

In September 2019, CATARC and ASAM jointly established the C-ASAM working group. Early member companies of the C-ASAM standard setting working group include Huawei, SAIC, CATARC Data Resource Center, Tencent, 51VR, Baidu, etc.