Looking back on 2023, most car companies are immersed in the keyword "intelligent driving", and "opening 100 cities by the end of the year" and "opening the whole country" have become the goals they are chasing this year. At the end of the year, when everyone was wondering which car company had achieved the goal and which one had been slapped in the face, car companies had successively released their L3 autonomous driving test licenses, announcing that in 2024, the goal of car companies would not only stay at intelligent driving, but L3 autonomous driving in the future stage.
However, the current L3 autonomous driving is still in the testing stage, and there are restrictions on the use scenarios and vehicle operating speed, so it will still take some time before large-scale commercial implementation. However, the test model information disclosed by various car companies has made it clear that L3 autonomous driving models need to strengthen their positioning accuracy and participation in the road, and the hardware and software must be adjusted to a certain extent. It will be interesting to see what parts of existing models need to be modified to comply with Level 3 autonomous driving standards, and how much money will be available for the first commercial Level 3 autonomous driving.
Yesterday, AITO announced on the last working day of 2023 that the M9 has obtained L3 autonomous driving test licenses in Chongqing and Shenzhen at the same time. On December 27, BYD also disclosed that the company had obtained the country's first high-speed deer conditional L3 autonomous driving test license as early as July. Combined with the previous relevant news, there are 9 car companies with test licenses nationwide, namely Cialis, BYD, Changan, Zhiji, GAC Aion, Beiqi Jihu, Lotus, BMW and Mercedes-Benz.
However, it is worth noting that most of the L3 autonomous driving test models of various car companies are active models on the market. For example, AITO Automobile's Wenjie M9, Changan's AVATR 11, Lotus Eletre, Zhiji LS6, etc., this means that the L3 autonomous driving of each car company will be based on the existing models to make software and hardware changes, which can directly meet the new autonomous driving standards, rather than starting from the test models that are almost "showing off their skills".
The modification of Level 3 autonomous driving of existing models is not an easy task for car companies, which means that the perception system and decision-making and judgment model in the car need to be overhauled to meet the requirements of the autonomous driving level. In terms of hardware, the Mercedes-Benz DRIVE Pilot system incorporates components such as wheel humidity sensors and rear cameras, allowing the car to perceive more information about the road environment to optimize the judgment logic of AI artificial intelligenceThe M9 and Deep Blue solutions also add 4D millimeter-wave radar on the basis of the existing hardware scale, which is used to improve the detection ability of object height and improve the accuracy of judging the surrounding environmentEven Lotus's solution adds four lidars to enhance the detection of surrounding obstacles. Whether it is Mercedes-Benz, Wenjie, or Deep Blue solutions, strengthening the perception accuracy is the main idea of optimizing the hardware.
The Mercedes-Benz Drive Pilot system brings us a new idea by adding a rear camera and microphone to monitor oncoming cars from behind. If a car from behind sounds a siren, AI will judge it as a high-priority vehicle and the vehicle will take action to avoid it. Compared with the current mainstream visual radar obstacle recognition scheme, Mercedes-Benz realizes the identification of vehicles with different priorities through a simple set of equipment, which also fully proves that the existing L2++ intelligent driving hardware can meet the needs of L3 autonomous driving after adjustment.
In addition, changes in software are more likely to start from the decision-making model. Because Level 3 autonomous driving requires a higher level of vehicle engagement than a human driver, the system needs to continue processing even after a feature request has been made until a human takes over. Therefore, the difference between L2++ and L3 is that the system needs to deal with more special situations on the road and achieve a complete closed loop of driving logic, rather than abruptly exiting like the existing intelligent driving system and asking humans to take over at all times.
It can be inferred that more processing logic will be added to the changes in the decision-making model of L3 autonomous driving models. The current L2++ level intelligent driving is more similar to the machine idea, that is, "find the problem - make a decision - the decision cannot be run through - the function exits", the problem caused by this logic is that the intelligent driving system encounters any problems after the processing method is taken over by humans, resulting in sudden program exit. The logic of L3 autonomous driving should be changed to "find the problem, make the corresponding decision, judge whether the decision can be run, notify the human to take over, no takeover, pull over", and use a complete logical closed-loop processing, rather than the function activation in the L2 era.
Therefore, we can infer that the main change aspect of L3 autonomous driving lies in the iteration of the perception and decision-making module, so that the vehicle's perception system and decision-making system are more similar to human models.
Although brands such as Wenjie and Changan have not announced the details of their L3 autonomous driving test models, according to the "Notice on Carrying out the Pilot Work of Intelligent Connected Vehicle Access and Road Access" issued by the Ministry of Industry and Information Technology and other four departments in November, the first premise of the autonomous driving plan is to ensure safety, which also requires car companies to improve the decision-making model to ensure the safety of road tests.
It can be seen that next year, car companies are likely to fine-tune the existing hardware to strengthen the perception ability, and the most difficult part is to optimize the decision-making logic of the autonomous driving system. **At this stage, the purpose of L3 autonomous driving road testing will focus on using the optimized large model to collide with the existing environment and assist the iteration of the autonomous driving large model.
Similarly, at this stage, the L3 autonomous driving standard is still in the revision stage, and various car companies are also accumulating specimens for the development of standards. It is expected that the L3 autonomous driving standard will be introduced in June 2024, and the commercial L3 autonomous driving models will also be launched in Q3 or Q4, which may be the fastest progress of L3 autonomous driving.
If the commercialization of L3 autonomous driving can be realized next year, how many will the first batch of options be reached?Perhaps according to the conjecture of the current test model, the optional ** of L3 autonomous driving ** may not be cheap, and even the minimum will reach about 30,000**.
First of all, the current hardware solutions of L3 autonomous driving are still focused on high-definition map + perception hardware stacking, so the hardware cost will be increased to a certain extent. Taking AVATR 11 and Wenjie M9 as examples, the two L3 test models are equipped with multiple lidars and 4D millimeter-wave radars, but the cost of lidar is about 1-20,000 yuan, and the single 4D millimeter-wave radar is maintained at the price of 1500-2000 yuan, plus the annual cost of 500 yuan for high-definition maps, the hardware cost is really not low. If L3 autonomous driving hardware is equipped with 3 lidars and 4D millimeter-wave radars as standard, the overall cost will increase to nearly 30,000 yuan.
Therefore, there is a high probability that various car companies will still adopt the current model configuration scheme of the M7, that is, the L3 level autonomous driving hardware is included in similar "intelligent driving version" models, and the model ** has a certain gap with the existing intelligent driving model, so that early players can pay for the experience.
In terms of specifics, BMW's L3 autonomous driving solution is available for 6,000 euros (about 4.) in Europe670,000 yuan), Mercedes-Benz's drive pilot system optional cost is divided into 5,000 euros and 7,430 euros two gears, about 350,000-520,000 yuan. However, Mercedes-Benz's L3 autonomous driving solution is inferior to domestic car companies in terms of hardware scale, and the hardware cost of domestic car companies' L3 autonomous driving solutions is higher than that of Mercedes-Benz BMW.
However, judging from the current intelligent driving pricing scheme, most car companies will choose the software option package service for L3 autonomous driving, similar to Huawei ADS 20 pricing logic, hardware cost is covered by the price of the model, software cost is covered by the optional package. There is a high probability that the L3 autonomous driving model will be 3-50,000 yuan more expensive than the existing version (the same as BMW and Mercedes-Benz), and the software service package will maintain an optional cost of 1-20,000 yuan. If users want to experience L3 autonomous driving, they will have to pay at least about 50,000 yuan more than the current intelligent driving model**, and the price of the software optional package will increase with the rich functions.
In 2024, it will enter L3 level autonomous driving, although the function may not be perfect, but it may be the cheapest year for the option.
The L3 autonomous driving model in 2024 may not be a very practical function for users, and the L3 will still experience embarrassing things encountered when the L2 intelligent driving system is popularized, such as identifying the wrong obstacles, and the system automatically making strange decisions and judgments. However, L3 autonomous driving is a key point for car companies to achieve differentiation next year, just as they are competing for test licenses this year. The only suspense now is that several car companies will enter the L3 commercialization track by the end of next year, hoping that L3 will no longer be a concept of show-off.