news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

the first half of the automobile industry is about electrification, and the second half is about intelligence. this has become an industry consensus. the discussion on intelligence has become increasingly heated with the rise of new energy vehicles. for example, the hotly debated #can traditional fuel car companies make smart cars? # quickly topped the best-selling list, which is based on the discussion of the development of new energy intelligence.

the reason why this topic has sparked discussion among many people is that, based on the current development trend, there is indeed a gap in the overall intelligence between fuel vehicles and new energy vehicles. in addition, traditional fuel vehicle companies have not given many surprises to consumers in the field of smart cars, and are not as good as new forces in the dimension of smart experience that the industry pays attention to.

on the other hand, the influence of tesla, hongmeng intelligent driving, nio, li auto and others in the intelligent field is increasing day by day. they can be regarded as the driving force behind the popularization of smart cars, and their influence on intelligence is naturally higher than that of traditional fuel car companies.

as well as the promotion and exaggeration of marketing publicity, the resulting publicity effect is unbalanced, which easily conveys the message to the outside world that traditional fuel vehicle companies are far behind new forces or new energy vehicle companies in terms of intelligence.

is it true that traditional fuel car companies are not good at making smart cars, which is the fundamental factor for the development of the domestic auto market, or is it that the general public's understanding has long been blocked in the information cocoon woven by internet marketing? in the discussion on this topic, many executives of auto companies also expressed their views.

among them, lu fang, ceo of lantu auto, believes thatit is impossible to make a smart car based on traditional fuel vehicles.fuel vehicles have inherent shortcomings in terms of intelligence, such as long system response time and insufficient redundancy, which leads to slow progress in intelligence in fuel vehicles, or in other words, fuel vehicles are not the best carriers for intelligence.

saic volkswagen executive fu qiang once said at the passat launch conference: "the intelligence of passat pro surpasses luxury brand fuel vehicles and is on par with mainstream new energy vehicles."it is believed that "the intelligence of a car has nothing to do with the form of power, and the intelligence of a car must be decoupled from the form of energy."it can be seen that automotive executives have different understandings of intelligence.

what is the core point of intelligence?

back to the level of smart cars, when we discuss the intelligence of cars, we generally refer to smart driving and smart cockpits. when it comes to smart driving, it is indeed difficult to build a smart car based on the traditional fuel vehicle architecture.because the electronic and electrical architecture of fuel vehicles is mainly based on the public type, the vehicle's power, chassis, electrical, cockpit, and intelligent driving are basically independent of each other.if fuel vehicles want to achieve a fairly good level of intelligent driving or smart cockpit, they will need higher cost investment than new energy vehicles, which is obviously not in line with business logic.

however, when it comes to intelligent driving of fuel vehicles, the models with outstanding intelligent driving capabilities in the traditional fuel vehicle camp include passat pro and xingtu lanyue.the two models are equipped with iq. pilot intelligent driving and noc automatic navigation assistance.their intelligent driving is considered to be the best among traditional fuel vehicles, but when it comes to new energy vehicles, there is still a gap compared with huawei, xiaopeng and other intelligent driving companies.

traditional fuel vehicles are slightly inferior in terms of intelligent driving, which ultimately comes down to the gap in electronic and electrical architecture. if traditional fuel vehicles want to do a good job in intelligent driving,it is necessary to reconstruct and innovate the perception system, information transmission system, computing system, and control system.

under the guidance of the "new four modernizations" leading the transformation and upgrading of automobiles, the electronic and electrical architecture of automobiles has gradually evolved from distributed to domain control and centralized, the intelligent control of the entire vehicle has become more centralized, and the weight has gradually shifted from the hardware to the software level.

more and more car companies such as saic, volkswagen, ideal, volvo, byd, great wall, and nio have entered the domain control architecture stage.xiaopeng even used it on the xiaopeng g9central supercomputing + regional control,intelligent data centralized processing enables more precise and faster control.

therefore, when we understand the structure of smart cars, we find that traditional fuel vehicles are indeed not the best carriers. traditional fuel vehicles have delays in power output. smart driving emphasizes safety, and millisecond-level delays may exponentially increase the risk factor at critical moments.

the motor drive of new energy vehicles can achieve extremely low latency and can also achieve rapid response and adjustment at high speeds. the extremely fast response capability required for intelligent driving will directly determine the strength of intelligence. in other words, fuel vehicles can achieve good intelligence, provided that the electronic and electrical architecture is reconstructed.

why is end-to-end intelligent driving so popular?

with the development of intelligent driving, xiaopeng, ideal, huawei and others are betting on end-to-end intelligent driving. whether it is a new energy vehicle company or a traditional fuel vehicle company, the competition for high-end intelligent driving of smart cars may be brought back to the same starting line due to the competition of end-to-end architecture. end-to-end intelligent driving technology requires the use of ai large model training.one end of the system needs to input perception data (camera, lidar, navigation information, etc.), and the output end directly gives decisions (driving trajectory). the biggest advantage of the integrated end-to-end system is that it is more efficient and intelligent.

at present, the intelligent driving systems of most car companies basically set the driving operating specifications in advance on the program side. when encountering a certain road, it will make comprehensive calculations and control the vehicle driving according to the preset rules.the disadvantage is that when encountering some unseen road conditions, such as a stopped vehicle at an intersection, the system may crash or wait indefinitely.(cannot be identified or the scene is not collected at the system level), manual takeover is still required in the end, and the vehicle will not drive in other lanes.

end-to-end training will be done through massive amounts of data, which is equivalent to equipping the vehicle's intelligent driving with a "brain". by learning various road conditions in daily driving, it then calculates and analyzes to give reasonable driving operations.even when it needs to drive in the opposite lane, it will automatically turn the steering wheel without hesitation, which is very close to the thinking of a real person driving.

of course, in theory, if you want end-to-end driving to be more like a human, the prerequisite is to provide sufficient data for ai large model training and iteration, making intelligent driving operations smarter, and even achieving l4 and l5 autonomous driving.

when facing the end-to-end architecture of tesla fsd, yu chengdong once said that the upper limit of fsd is very high and the lower limit is also very low. if the end-to-end system is not trained enough, when it cannot identify the obstacle in front, it may directly hit it without slowing down.

therefore, returning to the topic discussed in the article, if traditional fuel vehicle companies construct smart cars based on oil vehicle frames, it will obviously be difficult for them to surpass new forces in intelligent driving.

nowadays, as the end-to-end architecture opens up a new era of intelligent driving, it is believed that this is an opportunity for traditional fuel vehicle companies. even by collecting a large amount of driving data and combining it with ai large model training, the intelligent driving system can complete its self-evolution. with the increase of driving data, it is not impossible for traditional fuel vehicle companies to catch up with new forces in intelligent driving.

therefore, for any car company, whether it can make smart cars or not, although fuel vehicles are not good carriers, it is a false proposition that fuel car companies cannot make smart cars. with the use of end-to-end architecture and combined with the training of large ai models, it is not difficult to achieve equal rights for intelligent driving.