Shulou(Shulou.com)11/24 Report--

On August 11, Beijing time, California, the most conservative region in the United States, held a hearing on whether driverless taxis can be legally on the road. After a six-hour debate, self-driving supporters won by an overwhelming 3:1 vote.

Ride-hailing company Cruise CEO Kyle Vogt announced on the spot that thousands of L4 self-driving taxis will be launched in San Francisco in the next six months to provide 7 × 24-hour service to local residents.

(photo Source: Cruise) however, a week after the ban on self-driving taxis was lifted, the California Motor vehicle Administration regretted it and immediately ordered strict control of self-driving taxis. In just one week, what happened to the driverless taxi that made the California Motor vehicle Administration so angry?

In a week, self-driving is driving California residents crazy. In order to persuade the California government to allow self-driving taxis on the road, Cruise produced a lot of data to prove that self-driving is safer than manned driving.

Cruise driverless taxis have had many accidents since they were on the road in California for a week. Let's take a look at some typical cases.

First, at 10 p.m. Last Thursday, a driverless taxi in Cruise came into close contact with a fire engine at an intersection. The good news is that the accident was not serious. Under the protection of airbags, the passengers were not seriously injured and were taken to hospital in time.

In this regard, Cruise's explanation is that the car has detected, identified the fire engine and taken braking measures, but failed to avoid the accident because it was at a crossroads.

(photo Source: Cruise) well, I think you have a reason for this, but what about this one?

On the same day, a Cruise driverless taxi plunged into the road under construction and was "welded to death" by cement. After all, AI is not a real person, it is difficult to tell whether the cement is dry or not, but the taxi company is still inextricably linked.

Of course, the main responsibility lies with the road administration and road construction companies for not placing isolation facilities at the road construction sites. The responsibility of taxi companies is that the map does not collect data on road construction. You know, almost all online map platforms in China collect data in real time. In addition to official surveyors, users can also upload road information.

Since we are determined to develop autopilot, this should not be taken into account.

The biggest problem caused by Cruise was last Friday, when ten cars suddenly stalled at the intersection, causing the whole street to be blocked for more than 20 minutes. The reason given by Cruise is that a local music festival is being held, disrupting the signals of driverless taxis and causing the vehicles to lose data.

This is even more outrageous. Do cars have to be connected all the time? What about going through the tunnel? More than 60% of both 4G and 5G base stations are in China, so China cannot guarantee that there are signals everywhere. The coverage of base stations in the United States is much lower than that in China, even if the signal is not good. I am afraid of interference from other signal sources. I really don't know what Cruise thinks.

On August 18, the California Motor vehicle Administration, which could no longer stand it, ordered that Cruise in the San Francisco area should not operate more than 50 driverless taxis during the day and 150 at night, and launched an investigation into Cruise, which could disqualify the taxi company from operating self-driving taxis.

Waymo, another taxi company, has not exposed many accidents this time, but it has had many problems before, such as the first driverless car hit to death in 2018 and dozens of collisions with fire engines.

Within a week, autopilot exposed a lot of problems, many of which were related to safety. In recent years, many countries around the world have blown the horn of commercial autopilot, we can see so many problems, how can we rest assured to use autopilot?

Autopilot into the future, there are still many difficulties to overcome the speed of tens of kilometers per hour cars galloping on the road, the slightest carelessness may be killed in a car crash, we naturally pay more attention to.

To achieve autopilot, first, we need to collect data through the sensor, and secondly, we need the chip to combine processing and give instructions. From the perspective of security, naturally, the more sensors, the better, and the more accurate the data collected, the better, but a large number of high-precision sensors will cause greater pressure on the chip, not to mention the cost.

After entering the automotive industry, Intel, Nvidia and Qualcomm are developing high-power chips, of which Nvidia's Thor chip has reached 2000TOPS (2000 trillion operations per second).

(photo Source: Nvidia) in terms of radar, ultrasonic radar and millimeter wave radar are more popular, and some cars with less than 150000 yuan will be carried. Lidar is expensive, and only some medium-and high-end models can carry it for the time being. Domestic companies such as Huawei and Wesai are studying how to reduce the cost of lidar and strive for opportunities for low-end models to be used.

There are also some car companies who think that radar is superfluous and that pure vision solutions are enough to achieve autopilot. Humans do not have radar and rely on two eyes. For example, Tesla, the new Model 3 is about to be released, cutting off all radars. However, the human body is far more complex than any machine or AI, and the brain supplement ability of the brain cannot be matched by computers, not to mention that we collect environmental information not only with eyes, but also with many organs responsible for collecting auditory, tactile and other information.

(photo Source: Tesla) what's more, no car company uses a pure radar scheme, either pure vision or camera + radar fusion vision. Pure vision scheme has extremely high requirements for the algorithm, but can have radar autopilot scheme, can not ensure safety, let alone pure vision + algorithm, Tesla autopilot accidents are not uncommon.

Some relevant domestic enterprises have realized that it is impossible to achieve L5 autopilot only by relying on sensors, no matter how excellent the algorithm is. As a result, these enterprises have launched high-precision maps, vehicle-road coordination and other programs. Among them, the high-precision map needs to be accurate to the centimeter level by constantly surveying and mapping road data, so that there is no need to collect road information in real time to reduce the pressure on the chip.

Vehicle-road coordination is to add cameras in areas with complex road conditions and traffic congestion, collect information that can not be detected by car cameras and radar, and then send it to cars, so as to improve driving safety.

However, no matter which plan, you can see an obvious problem, that is, a lot of cost. Tesla Model 3 wants to cut all radars because he wants to reduce costs. Vehicle-road coordination and high-precision maps all need time, manpower and material resources to improve continuously, so it is difficult to achieve national coverage in a short time.

The future of autopilot is bright, but today, it cannot dominate the auto industry.

Autopilot, should be put back in the cage only to watch online publicity, it seems that autopilot has matured, just wait for the relevant departments to say, can be immediately commercial. In fact, now commercial autopilot only uses some consumers as experimental subjects.

As for whether autopilot should be opened up for commercial use now, Xiaotong believes that it still needs to be carried out step by step, and it is not advisable to act in undue haste. At present, many car companies have launched high-level auxiliary driving, such as Huawei ADS and Xiaopeng XNGP, which have performed very well.

Perhaps high-level auxiliary driving is a function that consumers should pay more attention to today than autopilot, which may be exaggerated. Autopilot is promising in the future, but it is best to be put back in the cage at present. After all, everyone's life is precious and should not be tested by self-driving companies and car companies.

The road test of self-driving vehicles in China is progressing steadily. Beijing, Wuhan, Guangzhou and many other cities have self-driving online car-hailing, which can only be operated on a small scale for the time being, and the price is very cheap. Everyone can also experience it if they have the opportunity.

When these car companies collect enough data to ensure the safety of consumers and are willing to take responsibility after the accident, autopilot can really enter thousands of households.

This article comes from the official account of Wechat: ID:dianchetong233, author: lost Soul.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.