"The development of smart cars is not because of how advanced the technology is, but to solve the problems in China's development, such as traffic safety and efficiency. And these problems are solved through automatic vehicle driving or an orderly system. We must consider fairness and whether the whole society can share such results." On June 20, Wu Wei, Director of the Industry Coordination Department of the National Development and Reform Commission, said at the GIV2018 Global Smart Car Frontier Summit.
In recent years, with the rapid development of the electric vehicle industry and the continuous breakthroughs in intelligent network technology, the development process and route of intelligent vehicles have attracted much attention. Among them, autonomous driving technology has become an important part of it. However, to implement this technology, future application scenarios How to build it requires exploration by companies in various fields.
Therefore, the selection of scenes is particularly critical for the research of autonomous driving technology. In this regard, Xia Huaxia, general manager of the Unmanned Distribution Department of Meituan, said: “Nowadays, many autonomous driving companies are struggling to find scenarios, but there are no landing scenarios. The reason is that autonomous driving requires a lot of maintenance, testing, and drivers. , And monitoring personnel, Meituan has no problem with the application based on the existing team, but in the automotive industry chain, Meituan cannot do it, but hopes to cooperate with other companies."
The challenge of the test scenario
At this forum, industry experts put forward the following challenges for the test scenarios: First, how to build the autonomous driving scenario library, whether to cooperate or build by yourself; second, the current cost of establishing a test site is too high, how to reduce costs and purchase Localization of the equipment; the third is to be applied after testing, from public transportation to logistics to actual social transportation, how to solve legal and standard issues to construct a travel ecosystem.
At present, most self-driving cars are still in the R&D and testing stage, and are different from traditional automobile technology, even in the sales stage, there will be continuous testing and evaluation. Sun Yafu, vice president of Qianfang Group, said: “The core of the current test is how to increase the cognitive ability of the vehicle and the learning scenario from the test field to the open road test, instead of requiring the vehicle to be used on the road after the vehicle is on the road. The standard for testing."
Of course, not all self-driving cars will be driving on the roads in the future. The situation of mixing self-driving cars and human-driven vehicles will continue for a long time, which will bring great challenges to road safety. Sun Yafu said: “The core level of how to ensure safety is to comply with existing traffic rules. A very core question for autonomous driving R&D companies is whether the R&D phase is to do R&D and design in compliance with traffic rules. The company is promoting autonomous driving. Cars move from closed fields to open roads to minimize the impact on existing traffic."
At the same time, he also mentioned: “At present, we are doing test services for more than ten manufacturers in Beijing. We have issued licenses for three manufacturers so far, and two more will be issued in the near future. Basically, there are 1 or 2 companies every month. The speed of driving forward. In the test procedures we designed, it actually covers 28 US highway regulations, including most of Google’s more than ten tests. At present, in the Beijing test, no manufacturer has passed all the tests. , Because if all tests are completed, basically the vehicle has reached the L5 level."
Not only that, many experts on the forum also gave suggestions that autonomous driving needs to consider how to cultivate an independent Chinese intellectual property system, such as artificial intelligence chips, during the test period. At the same time, it is hoped that the country can enter the automobile and transportation industries. Participate in the design of, electronic control and artificial intelligence chips, and promote in-depth cooperation between vehicle companies and autonomous driving software development companies, component companies and integrated circuit companies.
“At present, we divide autonomous driving tests into three types: virtual testing, parallel testing, and field testing. Among them, in parallel testing, product output faces many problems when testing instruments, because millimeter wave radar, lidar and other equipment are not domestically produced. Yes, it’s difficult to get its internal interface, so it can only do pixel or physical layer signal injection, and there is no way to do algorithm injection. Therefore, only in cooperation with component suppliers, can the entire vehicle in the loop parallel test be completed The link." Geely University Professor Wang Jian said.
The delivery field may be the first to land
According to the current automatic driving technology solutions, there are three difficulties: one is the need for long-term and large R&D investment, which puts higher requirements on corporate funds; the second is to use an environmental simulation platform, which is very difficult to perform full-featured simulation; third, to find a suitable one Application landing scenario.
At present, in the field of autonomous driving, there are not only Internet, vehicle, parts and other companies, but also express delivery companies such as JD.com, SF Express and Meituan are all in the layout. Among them, Meituan responded most positively, proposing to implement unmanned distribution in the park this year and achieve regional large-scale operations next year.
According to Meituan’s plan, in the future, driverless cars will be mainly "small cars" indoors and "big cars" outdoors. At the same time, Meituan will work with some partners to solve different scenarios. Xia Huaxia said: "Based on our judgment on unmanned delivery vehicles, the efficiency of unmanned delivery vehicles will not be lower than that of humans in the next three to five years. Although they are not as flexible in interaction as the delivery staff, they have greater capacity for reprinting and working hours. Advantages. In addition, with the rapid decline in lidar costs, the cost of unmanned delivery vehicles will gradually decline."
At the same time, he also mentioned: “In the R&D stage, companies don’t particularly care about costs, but after mass production, Meituan can control it at around 100,000 yuan even if it launches a ‘big car’.
It is conceivable that as vehicles are gradually upgraded along L3, L4, and L5, autonomous vehicles will bear more and more responsibilities. Once a vehicle has an accident, how to ensure the safety of passengers needs to be considered by enterprises during the test. "No matter what stage the autonomous vehicle is in, the scene will always be an entry point, because only a clear scene will know how to develop in the future, and the construction or improvement process of the scene must be done jointly by the enterprise, and it cannot rely solely on a certain manufacturer or a certain one. The unit is completed.” said Gao Youcai, senior manager of Bosch's autonomous driving cognition and braking system.
下一条：Notice of regulating the use of corporate office computers
