In order to achieve the understanding of human driving and driving behavior of cars and roads, it is necessary to collect a large amount of data for analysis. The two companies that currently have the most data are Tesla and Waymo. According to The Verge report, the two companies collect data. Not the same. Tesla used hundreds of thousands of electric vehicles with Autopilot semi-autonomous driving systems sold to collect real-world data to understand the performance of these vehicles. Waymo, which is independent from Google’s automated driving program, uses powerful The computer simulates and feeds back learning results to a relatively small number of real-world teams. Morgan Stanley analyst Adam Jona believes that for Tesla, the value of the data may be higher. Currently Tesla is passing through. Hundreds of thousands of owners around the world gather performance information on Autopilot functionality. In addition to mileage, Tesla also collects data on how Autopilot handles different driving scenarios when semi-autonomous driving is not used. The Tesla car records examples of actions the Autopilot software wants to take. , pass the data back to Tesla and let the company simulate all the billions of miles Autopilot data. As for Waymo, in early 2018 it announced that it had simulated 5 billion miles of autonomous driving data and reached a milestone of 5 million miles driving on public roads. Waymo's limitation is that it can only be collected by about 500 to 600 self-driving teams. Real-world data, and these vehicles are only driven in states such as Texas, California, Michigan, Arizona and Georgia, and the diversity is not as good as the Tesla electric vehicles with more than 300,000 in the world. However, Waymo plans to add thousands of people by the end of the year. Vehicles, more recently announced the development of a fully self-drive version of the I-Pace SUV electric vehicle in cooperation with Jaguar Land Rover. Waymo said that the number of vehicles will increase to 20,000 in the next few years. Once these cars are on the road, the accumulated mileage will be more It is worth noting that the data collected by Tesla and Waymo is not only of a poor scale, but also of different content. Waymo uses 3 different types of LiDAR sensors, 5 radar sensors and With 8 cameras, the Tesla is equipped with 8 cameras, 12 ultrasonic sensors and a front-facing radar. Guangda can help build high-resolution panoramic images of the car's surrounding environment. In the dark, accuracy can still be maintained, performance is better than cameras, radar and ultrasound, but Tesla is not used. Collecting data is one thing, and data processing is also a difficult task. In this regard, Waymo is obviously more than Tesla. Advantages. Waymo built a complete computer model of the test city, rebuilt real-world driving data on a computer with 25,000 virtual cars, and then uploaded the learning experience to the test vehicle. The simulation test required a lot of investment, resources, time And research, Waymo has a parent company to provide support, the current simulated mileage is close to 6 billion miles, and plans to launch a commercial taxi program at the end of the year. However, in the self-driving race, NVIDIA is none other than NVIDIA's technology is being Hundreds of companies including Tesla use it. In March, NVIDIA introduced the DRIVE Constellation simulation system to make it easier for companies without Tesla and Waymo fleet size or financial support to enter the self driving area. Miles for road testing and simulation The number is not enough to determine which company has the ultimate advantage. Another problem is how to define the safety of self-driving cars. The number of so-called lift car (disengagement), according to track the safety of the driver must regain control of the car autopilot system with a universal index of all companies, but this index is not perfect, because the definition is not precise enough, it is easy to fake.
