The Eve Of Commercialization Of Automatic Driving: Facing Technical Challenges And Ethical Dilemmas

With the rise of the new technology and industrial revolution, automatic driving, as an advanced form of intelligence and networking, has become a competitive position for the automotive industry. In recent years, this gorgeous circuit has poured into countless top technologies and talents, attracting countless capital and customers.

The emergence of AI will provide unlimited possibilities for future travel. With the gradual application of primary assistant driving function, autopilot is no longer a remote technology concept, but is becoming a real experience that is within reach.

However, high level automatic driving is completely far away from manual operation. Constrained by various factors such as core technology breakthroughs, cost control and standard regulations, high level automatic driving still has a long way to go.

What is the technical level of automatic driving? What are the core problems that need to be overcome in its commercial application? In November 20th, a high-end dialogue on commercialization of automatic driving was launched at the 2019 China Automotive Industry Summit hosted by the twenty-first Century economic report.

The dialogue was chaired by Zhang Junyi, a well-known investor. The participants in the dialogue included Deng Zhidong, director of the visual intelligence research center of the Institute of artificial intelligence of Tsinghua University, Zhu Jiangming, chairman of the Zero run car company, the head of intelligent driving of the commercial vehicle technology center of SAIC, Lai Jie, senior vice president of Wen Yuan Zhihang WeRide project, Zhang Ning, general manager of Guangzhou Xiao Ma Zhixing.

In this conversation, the guests discussed the scenes and patterns of commercialization of autopilot, including the continuously heated RoboTaxi (driverless taxis) in the Chinese market this year, and also talked about the technical challenges and ethical dilemmas in the process of landing. Outside the industry, in order to promote the true application of automatic driving, the industry also appeals collectively to form a unified, high standard and credible testing and publicity mechanism.

Scene and mode

Zhang Junyi: what do you think are the most breakthroughs in the field of automatic driving this year?

Deng Zhidong: more ICT (information and communication technology) giants enter the field of automatic driving, research and development of incremental ICT parts technology for the automotive industry, layout of automotive products in the whole scene.

Zhu Jiangming: the industry has regressed to the expectation of intelligent driving. It turned out that everyone was going to cross L3 to L4 immediately, but at least from the perspective of car companies, autopilot and intelligent driving were like a child.

Lai Jie: the smart heavy truck jointly built by SAIC, Shanghai port group and China Mobile will realize the first operation and formation of public roads.

Zhong Hua: Waymo started a truly unmanned passenger operation in the United States to completely remove the front row security personnel. Wen Yuan Zhihang also established the first RoboTaxi joint venture in the first tier cities in China in August. Wen Yuan Yue hang was the first to establish an automatic driving joint venture with a traditional taxi company.

Zhang Ning: the ecological understanding between the three drivers of autopilot, car and travel is clearer. This year, many cities in China begin to operate commercially and automatically drive and travel. RoboTaxi is also our starting point.

Zhang Junyi: RoboTaxi is now conducting pilot operation within a certain range. What is the expectation of commercialization? What are the conditions for landing?

Zhang Ning: simply speaking from the ability of the system platform, it is hoped that in the next five years, L4 level automatic driving will be made in a certain area. But the commercialized operation with real profit will take five to ten years to jointly polish the automobile industry and travel industry.

Zhang Junyi: now many travel companies have not yet made profits. When can a driverless taxi break even?

Zhang Ning: five years can achieve a balance, and five to ten years will have higher revenue. At present, the cost of automatic driving is relatively high, but it is still about half as long as the Moore's law. According to this estimate, our cost will only be controlled under the level of US $1-2, and the profit can be expected.

Zhang Junyi: ask the same question. Is the toll for a driverless taxi 2B or 2C?

Zhong Hua: Wen Yuan knows how to choose RoboTaxi. First of all, because this economic account can be counted, it can earn money. Now the cost of the taxi plus the five risk one gold account for more than 64% of our revenue. Our aim is to replace this part of the cost.

We are optimistic about the commercial landing time of RoboTaxi, and we can achieve breakeven in two or three years or so. The way of charging is consistent with that of taxis. We do not want to make taxi calls a process of re education and acceptance.

Zhang Junyi: some people think that autopilot is better than commercial vehicle. What does Lai always think?

Lai Jie: the scene of commercial vehicles will be landing sooner. There are industry needs for mines, wharves, closed roads and structured roads.

But in the process of promotion, there is always a problem. Customers will not buy our intelligent driving car. What he wants is service. Finally, commercial vehicles may also have an operation platform, and intelligent vehicle driving, and even behind suppliers, chip integrators together to create a production circle, truly realize intelligent driving business landing. Of course, this road is still very difficult.

Technology and ethics

Zhang Junyi: whether it's automatic driving or automotive application, the bottom is the driving of core technology and parts. In the next stage of developing autopilot in China, what are the technologies that we need to break through?

Zhu Jiangming: mainly from three aspects: perception, control and execution. The executive part is the bottom and directly deals with the lower part of the body. They need higher safety. In terms of controller, the general vehicle controller is completed by various car companies, and the sensor represents the sensor and conveys the sensing information.

From the perspective of the whole hardware system, AI in China is relatively advanced.

From the point of view of sensors, there are two factions in the industry, one is perception by lidar, the other two is direct vision plus millimeter wave radar, such as Tesla. I personally feel that the feasibility of Tesla's model is even higher when the cost and life of the lidar are not resolved, because the resolution of the camera and the ability to pass through the rainy days are not worse than that of the lidar. In the next few years, lidar can also play a supporting role. These are good directions for development.

Zhang Junyi: the development path of automatic driving at home and abroad is different. Foreign countries emphasize bicycle intelligence. China has put forward the concept of vehicle road collaboration. Some people say that autopilot will fall faster when combined with vehicles. What does Zhu Zong think of it?

Zhu Jiangming: it is a mode to inform the owner of the road information by means of V2X fast connection. But I think this is only a supplement. It's hard for you to cover all the roads, because the cost is very high. In case of which part of the road is not covered, you will have to switch to manual driving, so the efficiency will be even lower.

Zhang Junyi: everyone who drives a car will encounter a dilemma. It is an ethical debate to encounter the danger of maintaining the lives of passengers or protecting the safety of pedestrians. What does Professor Deng think?

Deng Zhidong: the autopilot industry has not yet landed. It may not be serious enough to discuss this issue in depth. In fact, we can define responsibility and give priority to protection subjects through laws and regulations. From a technical point of view, the truly able to achieve L4 level of driverless cars, we can say that there is no such security problem.

Recommendations and requests

Zhang Junyi: whether it is intelligent network or automatic driving, it involves policy and legal supervision, especially when the market is immature. As a practitioner, what kind of advice or request do you have for the regulatory authorities?

Deng Zhidong: it is suggested that the official traffic department of the government shall issue the official automatic driving mileage and the "detachment" report to the public on a regular basis every year, rather than the road test enterprises themselves. When technology and time are ripe, road test licences and trial operation licences of "no safe drivers" can be tried. Of course, the premise is that the relevant government departments should formulate strict L2-L4 level safety supervision standards and product production quality standards.

Zhu Jiangming: now automatic driving is really hot. There are many departments in China. We hope it is better to have a unified department to take the lead. There are many test zones in the national intelligent network, but some test areas are more of a form. There are not many truly automated driving test environments, which need to be unified from the national level.

Zhong Hua: driverless involves personal safety, safety is no small matter, for enterprises and the public is the primary consideration. It is hoped that the government will improve the entry threshold for automatic driving, not the threshold of money, but the threshold of quality. An autopilot "driving license" (automatic driving road test license) is still expensive in China. If you only take the money to make the threshold, it will not guarantee safety.

There is still a gap between the current test and the actual travel scenario. How to raise the threshold and hope that the colleagues in the industry can discuss it together and check the quality of the standards.

Another point is that the market is further liberalized, constantly opening up the scope of road testing and operation, and ultimately allowing the company to support itself. Therefore, it is our hope to raise the threshold and open the market.

RJ: my point of view is similar to that of Chung Chung. There should be a national functional test. Now, the standards of smart network testing have been launched everywhere. But after testing, can the vehicle be safely on the road? This is a big question mark.

Vehicles should be implemented gradually through long term operation and data accumulation. But now, the closed test scenarios and open roads are also good for testing purposes, not being combined with the actual application scenarios. There is a good road condition on the open road and no one has no car. This has no meaning for automatic driving.

The intelligent network Federation test road and test area should be combined with the actual application scenarios. If you want to complete the test, get the license plate, you can immediately put into the application scenario, such as logistics, bus and so on, which is more conducive to the landing of autopilot.

Zhang Ning: very much agree with the previous views. First, we need to raise the threshold. Of course, this is a feature with Chinese characteristics. There is no automatic driving test in California. As long as the certificate is complete and insurance is available, we can apply for the road. However, the two countries have different national conditions. China has the necessity of automatic driving test.

But increasing the gold content of examinations is very important. Our exam is still a simulation of driver's license, but it is a specific project driving skill, but less intelligent common sense considerations (generalization ability). With the accumulation of common sense and basic driving skills for decades, people can go on the road, but automatic driving is not the same. It is a question mark for the safety condition that has not been encountered before. Therefore, it is necessary to form a safer and more universal requirement.

The promotion of autopilot needs government support, but there is no set of national level standards to promote autopilot landing. We look forward to exploring more replicable models and standards with the support of laws and regulations.

Xiao Ma Zhixing has been strongly supported by Nansha District, Guangzhou. Nansha District's 800 square kilometers are open to Xiao Ma Zhixing. The maturity of our commercialization is measured by square kilometres instead of kilometers. To make travel services, it is necessary to meet the needs of ordinary citizens to drive a taxi, rather than to open only a few roads. If we only open more remote roads, or do not get through the areas, such a demonstration will be of little significance for RoboTaxi to enter the next stage.

?