Before issuing this question, we must first know that if a self-driving car really wants to go on the road, it must pass multiple safety tests, and road test is an essential and important link.
According to the estimation of the Rand think tank in the United States, an autonomous driving system needs to verify at least 1 10 mile (about170-1800 million kilometers) to meet the conditions of mass production. This means that even a self-driving fleet with 65,438+000 test vehicles will take about 500 years to test at an average speed of 25 miles (40 kilometers) per hour 24 hours a day.
At the same time, a mature self-driving car also needs to sensitively identify objects on the road and make correct and rapid responses under extreme conditions such as snowstorm, rainstorm and strong sunshine. However, in the natural environment, these extreme situations do not often occur, and many extreme situations are very dangerous and must be estimated in the test. Therefore, it is inefficient, costly and incomplete to only carry out self-driving road test on the spot. Under such circumstances, a simulation platform that can provide a perfect virtual reality road test environment has become a cost-effective choice for self-driving cars to carry out road test. Today, let's take a look at how those well-known autopilot simulation systems at home and abroad conduct virtual road tests.
NVIDIA: Drive? Constellation-a computing platform based on two different servers
NVIDIA, which is a world leader in programmable graphics processing technology, officially announced the listing of the cloud autopilot simulation platform Drive in March last year. Constellation
Driving? Constellation uses photo-realistic simulation and cloud-based self-driving car testing system, which is a computing platform based on two different servers. The first server is running? NVIDIA? Driving? Sim？ Software that simulates sensors (such as cameras, lidar and radar) of self-driving cars. Driving? Sim software can generate photo-level data streams to create a large number of different test environments. For example, it can simulate different weather conditions, such as rainstorm and snowstorm, light changes at different times of the day, and all different types of roads and terrain. The second server is equipped with? NVIDIA? Driving? Pegasus? AI？ The car computing platform can run the complete software stack of self-driving cars, and can process analog data like sensors on the road.
NVIDIA Drive？ constellation
Tencent: TAD? Simulation platform of Sim-like large RPG game
As we all know, many mainstream simulation systems of autonomous driving are developed according to game engines, and Tencent, which is good at game development and management, has also applied professional game engines, industrial vehicle dynamics models, virtual and real traffic flows and other technologies to the automatic driving simulation platform TAD? Sim (Tencent? Autonomous? Driving? Simulator), which is also the unique advantage of Tencent as an autopilot simulation platform.
At ted? Sim, which is infinitely close to the real world, can not only meet the iterative test requirements of autonomous vehicles, but also improve the research and development efficiency of autonomous driving technology. Ted. The high-precision map built in Sim can complete the closed-loop simulation verification of perception, decision-making and control algorithm of all modules in real vehicles, complete the geometric simulation under different weather and lighting conditions, and test the vehicle's perception ability, decision-making ability and vehicle control simulation. Combining the collected traffic flow data with the simulation of extreme traffic scenes, TAD? Sim can carry out all kinds of aggressive driving and extreme automatic driving tests, and complete all kinds of tests that cannot be carried out in the real world in a more efficient and safer way.
Last year, Tencent and BMW reached a strategic cooperation in the field of autonomous driving, analyzing and handling the complex and changeable driving scenes in China, and helping BMW to develop autonomous driving technologies and products in line with the China market. Besides, Ted. Sim can also provide tests for decision-making departments and traffic management departments in traffic dispatching management, road and traffic planning, and research on autonomous driving regulations. Fully assist the mass production of autonomous driving.
Tencent TAD? Sim complex road simulation
Waymo: Carcraft-the "pioneer" who turned reality into virtual reality.
From the initial "Google unmanned vehicle" to the later Google Waymo, Google can be said to be the "big brother" in the field of autonomous driving research. Not long ago, Google announced? Carcraft, a simulation test software independently developed by Waymo, has simulated the road scene of 1000 billion miles, which supports the large-scale test of Waymo models.
Since self-driving cars need to use network nerves and algorithms for continuous learning, various road conditions set by Carcraft can make vehicles learn more from them. Waymo？ Many situations encountered by test vehicles on the road can be directly blurred in the simulator, and programmers can also superimpose various situations to create various extreme situations. The data obtained from Carcraft's simulation can be fed back to the real-world test vehicle, so that the vehicle will become more and more powerful. For Waymo, who was still testing with "scene playback" before, Carcraft has played an unprecedented important role in the research of autonomous driving, and even for the world's autonomous driving technology, Carcraft is of great significance.
At the end of 20 19, Waymo? Announced the acquisition of Latent, a British analog technology company? Logic is used to help Waymo realize more realistic simulation technology and better predict and plan autonomous driving.
At present, the domestic autopilot simulation system is still in its infancy. Even companies with mature simulation technology in the world are not mature enough in domestic road scene development and driver behavior simulation. From another point of view, this has actually released a good opportunity for local companies in China to develop the simulation technology of autonomous driving. It is expected that more and more local scientific and technological enterprises like Tencent will join in the construction of automatic driving simulation system, build a simulation test environment with China characteristics, and help mature driverless technology to be realized at an early date.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.