Follow Tesla collectively and abandon high-precision maps. Is a purely visual solution the best solution?

Regarding intelligent driving and autonomous driving solutions, each brand has its own set of "theories."

Tesla has always insisted on taking a purely visual route. Although it has been questioned by many domestically, Tesla still needs a long development cycle to achieve the final expected autonomous driving.

Since Tesla launched auto pilot, various domestic car companies have also begun to crazyly lay out intelligent driving routes. Some brands have also followed Tesla's footsteps and adopted visual solutions, and even added many detection sensors on the basis of cameras.

Some brands are digging deeper into the adaptation of high-precision maps, but which route is the most reliable at the moment?

Are HD maps important?

In fact, the importance of high-precision maps has become a hot topic in the industry since around 2020. Both car companies and consumers themselves have their own arguments.

According to the perspective of domestic car companies in 2021-2022, domestic car companies have always believed that with the blessing of high-precision maps, the dual blessing of maps and vision can help to surpass pure visual solutions more quickly.

Therefore, the initial role of high-precision maps, from an objective point of view, is to overtake on corners.

However, as subsequent visual sensors become more and more advanced, the "noise" of high-precision maps seems to be much smaller. Why is this?

On May 2, Zhao Changjiang, general manager of Denza Motors, expressed his views on intelligent assisted driving on social platforms.

Zhao Changjiang believes that current smart driving cannot be called autonomous driving, but can only be called assisted driving; and the purpose of assisted driving is to make driving safer and more comfortable; Zhao Changjiang even believes that smart driving cannot rely on high-precision maps, and it is difficult to operate in urban road conditions. to real-time updates and accuracy.

Regarding the importance of high-precision maps, when Huawei launched the ADS 2.0 system, Yu Chengdong also said that this system will be less dependent on high-precision maps in the future, and said that it can be used even without high-precision map coverage.

This means that both Denza and Huawei will lower the weight of high-precision maps and enable deeper peripheral perception and control software development.

In addition to Denza and Huawei, Xiaopeng also adopted a deeper visual perception solution in the XNGP system version after launching G9. Even in areas without high-precision map coverage, Xiaopeng can also use LCC as the basis , and gradually add more capabilities, so that in areas not covered by high-precision maps, the performance of XNGP is infinitely close to that of NGP.

Therefore, judging from the current industry development trends, high-precision maps can indeed make smart driving more convenient and accurate. However, because the collection price of high-precision maps is too high and there is too much reliance on high-precision maps, pure visual accuracy cannot be guaranteed. Spend.

Therefore, judging from the methods of several cutting-edge companies in the industry, it is very likely that the route Tesla initially adhered to was correct.

How to continue to advance intelligent driving

There is no doubt that the ultimate route of intelligent driving must be towards autonomous driving.

Lei Jun also released a one-shot video of Xiaomi's self-developed autonomous driving in the previous exposure of Xiaomi's progress, and publicly stated that autonomous driving is the key to the success of smart cars in the future. It can be seen that autonomous driving is a "storage" for many car companies. Ready to go."

At present, for smart driving, regardless of laws and regulations, most brands can actually implement fewer functions. Almost all of them are based on LCC and can realize the function of turning lights to change lanes.

There are also fewer brands that can realize functions such as automatic lane changing and overtaking on the highway based on the functions mentioned above. It is also generally called an intelligent pilot assistance system by the industry and has stronger capabilities than LCC. , of course, there are still only a few brands that can realize this function.

There are also very few brands that can realize intelligent navigation assistance in cities, represented by Huawei and Xpeng, which have been opened to users in some areas. This means that intelligent driving is currently available in various car companies. There are uneven performance among them, and many of them are still relatively far away from the long-term development path of autonomous driving.

But the route of intelligent driving must be unswerving.

Of course, some brands that "give up on themselves" are also excluded. For example, Mr. Wang Chuanfu, the founder of BYD, has previously publicly stated that autonomous driving is the emperor's new clothes, which has also been highly discussed in the industry.

At present, although high-precision maps have been considered the most suitable "companion" for domestic autonomous driving, according to industry consensus, the cost of collecting information for high-precision maps is too high. If you want to map the entire domestic roads and environment All of them were collected successfully, but the difficulty is indeed too high. Therefore, the author believes that the transformation of various enterprises from relying heavily on high-precision maps to visual solutions is a relatively correct direction.

Of course, although domestic brands are shifting towards visual direction, they are still at the leading level in the entire automotive market in terms of hardware use, such as laser radar, millimeter wave radar, ultrasonic radar, cameras and other visual acquisition sensors, which will definitely be better. For Tesla’s pure camera solution, of course, the improvement and optimization of software algorithms are also crucial. Therefore, the author believes that if domestic car companies can polish the pure visual solution to the extreme and rely on better combination of software and hardware, they will definitely be able to combine special features. Slala went into the water.

For current independent brands that want to overtake in corners in the field of intelligent driving, there is no doubt that the current chips and visual sensors are already at the top level of the industry, but in terms of software, more room for optimization is needed, especially not to have models from a certain car company. , in the deserted open space, the screen still shows the terrifying phenomenon of people walking around.

write at the end

Autonomous driving must be an indispensable general direction of automobiles in the future. Brands that give up early will face being eliminated by the market earlier. The purpose of technological development is to make users travel more conveniently and comfortably. Although intelligent driving has been actively progressing , but it is still in the "initial" stage, so when autonomous driving can enter thousands of households, it will still take a long time.

Assisted driving is the foundation of autonomous driving, and it requires car companies to be serious and responsible.