Is L5 automatic driving of "stacking type" desirable?

Netease automotive industry 2021-08-06 09:13:34 阅读数:903

l5 automatic driving stacking type

The author has participated in many forums in the automotive industry , The guests on the stage are usually very gentle , Even if there is an opinion , They will also express similarities and differences in a very tactful way . But recently in Auto Byte At a driverless Forum , Several guests on the stage were tit for tat , The argument was red in the face , Even grab the microphone several times to interrupt each other . A fierce collision , Let the audience shout , It also further leads to thinking about the driverless technology route ——“ Stacking type ” Of L5 Is level automatic driving desirable ?

The cause of the matter is Robo Taxi company Auto X Released the fifth generation of all unmanned driving system Auto X Gen5, In order to realize real all unmanned driving ,Auto X Gen Equipped with the world's top vehicle specification level hardware and electronic and electrical architecture , There are... In the whole car 50 A high-definition vehicle gauge sensor . this 50 A high-definition vehicle gauge level sensor includes 28 individual 8 Megapixel car gauge camera , The sum of pixels per frame is up to 2.2 Billion ; The highest resolution in the world 4D Millimeter wave radar , The angular resolution reaches 0.9 degree ; as well as 6 A high-definition lidar , Super per second 1500 10000 point cloud imaging . in addition ,Auto X Gen5 The core computing platform Auto X XCU The computational power has reached amazing 2200 TOPS, It has also set a new high in China's on-board supercomputing power .

But this kind of “ Stacking type ” Technical solution , But by yingche Technology CTO Yang Ruigang's question .

Yang Ruigang thinks , First, it's not that the more sensors there are, the more valid data there are , Secondly, autopilot has entered the stage from model to mass production , An important prerequisite for mass production and popularization is that the cost is controllable , Instead of stacking at no cost .

“ My personal biggest feeling is to do mass production of automatic driving is wearing ‘ Inhibition Charm ’ Automatic driving of , In other words, we need to consider more than security , Not just comfort , We also have to consider the vehicle specification level hardware , Also consider functional safety , It is particularly important that we also consider the cost .” Yang Ruigang said :“ I don't know any self driving company , Can have a self-supporting team that does not generate profits , To test tens of thousands of teams , Especially with 50 If you have a sensor , I think that price is astronomical .”

Regarding this ,Auto X Founder Xiao Jianxiong retorted ,“ The data collected by very good sensors is effective data , And relying on data from assisted driving , The effect on you is basically the same as downloading some video of dash cam on the Internet , There is no difference .” For this reason , Automatic driving must be equipped with enough high-definition sensors and a computing platform with enough computing power .

Yang Ruigang further retorted , Only when the cost is controllable , Automatic driving technology can be popularized ,“ Because of mass production 、 After scale , The cost of sensors can change from hundreds of thousands to tens of thousands 、 Thousands of , We all know the scale effect . mobile phone , Cell phone, if everyone transmits cell phone , I believe the mobile phone is not 3000 yuan , I'm afraid it's 300000 yuan .”

Huawei smart car cloud service CTO Although Yu Jie did not participate in the debate , But he stressed that in the development of automatic driving , The importance of high-precision map and vehicle road coordination is becoming higher and higher , His views are similar to those of Ding Lei, founder of Chinese express .

Ding Lei once told the author ,“ No matter what kind of sensor , Can not solve the problem of perceptual blind area caused by environmental occlusion , Only install the sensor at the end of the road , To solve the problem of blind spots .” in other words , Completely driverless through the road 、 The integration of car and city , Instead of relying on bicycle intelligence .

版权声明:本文为[Netease automotive industry]所创,转载请带上原文链接,感谢。