The road of mass production of automatic driving is the same as that of fighting monsters and upgrading. There are always thousands of “roadblocks” on the road, such as safety challenges, immature core technologies, imperfect laws and regulations, moral and ethical disputes, etc., which have been on the road ahead of automatic driving for many years, making this technology, even now, can only see the dawn through a long chasm.
To this end, Rand Corporation, a famous us think tank, once estimated that it would take 11 billion miles for an L5 class autonomous vehicle to go on the road. This means that even a self driving team with 100 test vehicles will take about 500 years to test continuously at an average speed of 7 × 24 hours per hour at 25 miles (40 kilometers) per hour. In this case, it is obviously unrealistic to only rely on the field to carry out automatic driving road test.
As a result, simulation test is useful as a technical means that can help auto driving R & D enterprises to reduce physical test mileage, reduce R & D costs and improve R & D efficiency greatly.
Simulation test is simply to simulate the road test environment by means of sensor simulation, vehicle dynamics simulation, advanced graphics processing, traffic flow simulation, digital simulation, road modeling and so on, and establish a mathematical model of real static environment and dynamic traffic scene, so that the automatic driving vehicle and algorithm can carry out driving test in the virtual traffic scene, so as to realize reality in a short time. Hard to reach test mileage in life. Especially for some extreme scenarios that are basically impossible to test on the actual road, or the automatic driving test scenarios that are not allowed by law, such as the automatic driving test on the highway, can be quickly reproduced in the simulation platform, so that the vehicle enterprise or the automatic driving company can carry out virtual test in it.
For example, waymo, based on simulation testing, reached 5 billion miles as early as September 2018. In July 2019, the data reached 10 billion miles, doubling the time of 10 months. In addition, Huang Renxun, CEO of NVIDIA, once publicly stated that if engineers use simulation test system, they can complete 480000 km road test within 5 hours. Therefore, at this stage, many enterprises in the research and development of automatic driving technology use the combination of automatic driving simulation test and actual road test to verify the safety and reliability before mass production.
It is to see that simulation test is indispensable for automatic driving. In the past few years, technology giants such as Google, NVIDIA, Baidu, Tencent, etc. have spared no effort to build their own simulation platform. At the same time, a large number of new players have emerged in the field, such as AAI, 51world, cognata, panosim, parallel domain, righthook, and latent logic just acquired by waymo, which are all new forces in the field of automatic driving simulation test.
Undoubtedly, simulation test is a key technology in the process of automatic driving from technology maturity to product landing. As Guo Jishun, head of Intelligent Driving Technology Department of GAC Research Institute, said, automatic driving technology should put safety in the first place, virtual simulation test can maximize the coverage of scenes, and ensure the safety and reliability of mass production technology. Because there will not be enough tests without simulation, there will not be enough safety without sufficient tests, there will be no way to go on the road without enough safety automatic driving, and it is impossible to realize commercialization.
And with the importance of this technology to autopilot cars, its market potential is constantly emerging. It is estimated that the total scale of the international market for simulation software and testing will be about 10 billion US dollars in the next five years. With the implementation of commercialization, the industry will force the further improvement of simulation virtual testing technology, so as to help automatic driving realize commercialization faster.