[YesAuto Fresh Technology Interpretation] When it comes to technology that has received the most attention this year, I am afraid that it is none other than autonomous driving. Open road testing, practical experience, theoretical explanation, for a while, overseas and domestic auto shows, auto company brand launch conferences and technology days seem to have fallen behind others without the appearance of autonomous driving technology. However, the distance between autonomous driving and ordinary consumers is still far away, and one of the fetters is unremarkable maps, or high-precision maps.
-Who desperately needs a high-resolution map?
If you ask me, who needs a high-resolution map, then we may have been inconclusive in the last hour of discussion, but if we talk about who needs high-resolution maps most urgently, then the answer is simple. The one that needs it most is autonomous driving. technology.
At this stage, some manufacturers have already started public road tests, but a heavy snowfall at the end of last year caused many autonomous driving test vehicles to “lie down”, because their “eyes”-sensors had snow blindness. The reason for this result is that traditional sensors are easily affected by the weather. Of course, we don't want our car to go on strike due to bad weather in the future.
In severe weather conditions, the problem that the self-driving test vehicle faces is not only the inability to judge the road conditions, but more importantly, it needs to use the information collected by various sensors to create a three-dimensional map during the movement process to determine its own location and plan the route. To the destination. When there is strong rain or snow, the reflection characteristics of the road will change due to icing or snow, and the information collection effect of the sensor will be affected to a certain extent, which will eventually affect the construction of the three-dimensional map, which will lead to the unmanned vehicle being unable to locate itself Location, so that it is forced to park in a roadside safe parking space waiting for a human driver to take over. This is the weakness of sensors that collect road conditions in real time.
Compared with radar, laser probes, and cameras that are like the eyes of autonomous vehicles, what high-precision maps bring to autonomous vehicles is like a memory of the front. The vehicle radar sensor that is widely used now has an effective detection range of less than 300 meters. In extreme weather such as rain and snow, or when the detection object is covered by a moving vehicle, this number will be greatly reduced. High-precision map It is not affected by this situation, combined with factual road condition information technology, it can still bring excellent road condition information under bad weather conditions. Therefore, for manufacturers, high-precision maps are a double insurance for automatic driving functions.
-What is a high-resolution map?
I believe that for most people, maps are not a strange thing. In the geography class when we were in school, we used it to read, memorize, and remember the customs and customs on the other side of the ocean over and over again. So, what is the high-resolution map mentioned above? How is it different from the navigation map we are using in our car?
Today, the map on our car has changed from paper-like to electronic version. At the same time, the map information and version are updated faster, which brings users a more reasonable navigation path and more diverse selection logic. However, all these changes are only a process of quantitative change. Fundamentally speaking, they are still only tools to assist the driver to reach the destination.
Information such as traffic lights, lane lines, and warning signs is the basis for the autonomous driving system to make judgments. It is precisely because the computer will replace the human brain as the main body of decision-making, so this puts forward higher requirements for the accuracy of the map or the map accuracy. So, what kind of map do self-driving cars need?
The generals who marched and fought a hundred years ago needed a map containing detailed terrain, rivers and even vegetation. Their grandson instilled in their minds the “knowing oneself, knowing the enemy, knowing one’s enemy, but never ending a battle”. Only by understanding the surrounding environment as much as possible can they reduce the possibility of being ambushed. . Today, autonomous driving technology requires high-precision maps for similar reasons. The purpose is to learn as much surrounding information as possible, but today's cars need more information and more accurate maps. The high-precision maps required by future autonomous vehicles need to be based on the existing basis, including detailed road slope, curvature, number of lanes, lane types, and other data and attributes related to lanes and lane lines.
According to the information currently obtained from the manufacturers, for the semi-autonomous driving stage technology that has appeared on the market, the map accuracy of 5 meters can meet the demand, while the fully automatic driving stage requires an accuracy of 10 to 20 cm. To upgrade the existing meter-level precision products to the centimeter level, many difficulties need to be overcome. Among them, there are both technical problems and policy relationships.
-Who can make a high-resolution map?
At present, our commonly used electronic maps are provided by map suppliers. The familiar domestic map suppliers include NavInfo, AutoNavi, Kaylide, Tencent, Sogou, Baidu, etc., regardless of the length of establishment, they have passed the national qualification approval. The overseas part is the world of maps such as Google, TomTom, and HERE.
-Difficulties to move forward
Since high-precision maps are one of the key factors in the realization of autonomous driving technology, why are there still few high-precision maps on the market? The reason is that there are many obstacles behind it, and the biggest obstacle comes from the national level.
It is for reasons of confidentiality that China currently adopts an access mechanism for mapping. However, according to the news obtained at the Asian CES exhibition, the Chinese government has observed the urgent need for high-precision maps in the civilian market and is considering opening up domestic highways. Regional high-precision map surveying and mapping authority , so from this perspective, in the future, domestic autonomous driving technology may be the first to be allowed to be used on highways.
-Is it not enough to overcome difficulties?
High-precision maps are actually similar to computer operating systems and mobile phone operating systems. The completion of the collection and sale is only the beginning, and then the company needs to maintain it. For IT companies, it is necessary to continuously roll out patches to repair system vulnerabilities and speed up operation. For map suppliers, as the real-world road conditions are constantly changing, map resources need to be re-collected on a regular basis.
The re-collection of map resources is not successful. The re-collected information needs to be sent to our car. At present, the maps used in the car navigation system are mostly upgraded when the car is repaired in the 4S shop. The demand for map update speed of autonomous driving technology must be in seconds. Only in this way can vehicles have a judgment time equivalent to or even beyond humans. Therefore, the update of map resources may be upgraded online through cloud technology. Of course, this is Will put forward higher requirements on the network.
At the same time, when cars in the future really use high-precision maps, the massive amount of information is not comparable to today, so this puts higher requirements on the performance of the vehicle's processor. In addition, after multiple system upgrades, its share The storage space is bound to become larger, which also poses a big challenge to the system storage space. On the other hand, after the hardware part meets the technical requirements, the cost issue cannot be ignored. In the short-term visible future, improving hardware processing capabilities and reducing costs are also one of the factors that determine whether autonomous driving technology can truly become popular.
-What will it affect?
Although the future development of high-precision maps is so bumpy, the numerous demands in the civilian field still drive the continuous development of map suppliers. The needs of individual users are only one of them, and more large-scale enterprises are the real demanders of high-precision maps.
Finally, and most importantly, is the impact of high-precision maps on autonomous driving technology. High-precision maps can carry more road condition information. Therefore, for autonomous driving systems, it can obtain relevant information in advance, thereby reducing real-time analysis and calculations. The demand for data will undoubtedly have a positive effect on reducing the intensity of system operation.
On the other hand, autonomous driving technology continues to advance, and the processor and system judgment logic can achieve a high degree of intelligence, surpassing human response speed and judgment ability, whether high-precision maps are necessary or not may become a brand new question.
Full text summary :
At this stage, for autonomous driving, it can only be a semi-finished product without the assistance of high-precision maps. The autonomous driving test vehicle, which mainly relies on various sensors to monitor the road surface, is no different from a blind person touching the image if it encounters weather problems. For self-driving technology that strives to be foolproof, it is impossible to put all the eggs in the same basket.
Therefore, it will inevitably require another means of obtaining auxiliary information, and high-precision maps are an important part of it. Such urgent needs will inevitably promote the continuous cooperation between car companies and companies in other fields. Car manufacturers provide in-depth understanding of the development and research of the car itself, coupled with the map collection that map suppliers are good at and the accumulation of Internet companies at the level of big data and cloud computing. , And finally pass the technical difficulties required for popularizing autonomous driving.
Overcoming “obstacles” while moving forward and analyzing autonomous driving technology