From knock-off old-man's electric vehicles to industrial giant

Chapter 207 Unmanned Parking, One-Click Summoning of Global AEB: We've Developed Three Types of This Amazing Feature
Unlike Xiaomi.

vivo is happy to pay for this battery.

The two parties finalized their cooperation after only twenty minutes of conversation during their first meeting.

"Mr. Xu, we have no objection to the price, but we require that the inventory be at least one and a half months' worth, and that the quality and yield rate of the cells after mass production be consistent with each batch with a difference of less than ±1%."

After going over the technical details, Shen Wei was very satisfied.

As this is our first collaboration, our biggest concerns are supply inventory and yield rate.

After all, it's going to be used on a flagship phone.

The samples that Xingchen has provided so far are of very high quality, but there must be follow-up guarantees.

As I said before, a little trust is needed for the first collaboration, and a single supplier also requires both parties to bear more risks.

Although the performance and lifespan of this battery are currently leading in the industry, there are no other alternative suppliers.

If StarCraft's production capacity encounters problems, it could potentially disrupt the entire flagship phone lineup. To avoid this, the only solution is to stock up on more inventory.

"President Shen, Xingchen is a professional in battery manufacturing." Xu Yi didn't say much, but showed a very confident attitude.

Quality and quality control have always been Xingchen's strengths. Now that the production capacity has been confirmed, the mobile phone battery cell production line, which has been laid out in advance, can start production directly. According to the agreement between the two parties, the initial inventory will be at least 150 million units.

……

mid-December.

Xu Yi was much more relaxed than before. Several prototype versions of the Starry Sky L7, after optimization, were successfully launched. After detailed testing and improvement, the hardware technology was successively verified.

Encouragingly, Xingchen's self-developed intelligent driving system and brand-new system software have officially entered a mature stage.

The testing ground at the R&D center.

Several engineering prototype vehicles equipped with a complete set of calibrated intelligent driving hardware are being tested inside the site.

In addition to the lidar mounted on the roof of one of the prototype vehicles to collect multimodal data.

The other vehicles all use mass-produced cameras and millimeter-wave radar sensors.

The overall configuration of the vehicle is almost identical to that of the mass-produced version; the only missing elements are tuning and minor adjustments.

"Mr. Xu, our self-developed Super Sensing ACC Cruise Control system has completed testing. Based on the NVIDIA DRIVE PX 2 platform, it processes the image data acquired by external sensors to create a high-precision point cloud. The system uploads all point cloud data from the DRIVE PX 2 platform to the cloud server, and after processing by DGX-1, it can be fused into a complete simulated map, capable of recognizing all highway conditions and road signs..."

Wu Xinzhou said.

The internal intelligent driving conference has been held multiple times, and progress is being made rapidly.

After the team and the intelligent driving center were set up, the ACC and LCC solutions were fully implemented, and the internally developed algorithm model has also made progress.

"Three teams are testing the highway sign and map data using test vehicles. We have now opened two main highways, covering the branch highways in the Shandong area. We expect to open more than 10 core national trunk highways before October."

Internally, the current intelligent driving project is called Navigate on Autopilot, which is L2+ level navigation assistance driving.

The primary routes to focus on in the early stages are the expressways centered around the Yangtze River Delta, the Pearl River Delta, and the Beijing-Tianjin-Hebei region. We will cover all the regional road sections connecting first-tier and new first-tier cities, including national trunk lines such as Beijing-Shanghai, Beijing-Hong Kong-Macau, and Shanghai-Chengdu, where high traffic volume means "high user demand," making them our main targets.

After internal discussions and Xu Yi's intense "proposal".

Xingchen Automotive's algorithm abandoned the high-precision map solution and directly started training a high-speed NoA algorithm without a map.

Of course, the initial no-map noa does not mean that an internet connection and a high-precision map are not needed.

The necessary road sections still require detailed data collection, combined with high-precision positioning data from navigation systems, to be opened.

Once their internal intelligent driving model is fully perfected, and model training, data annotation, and real-time decision-making on the vehicle are all completed, they will be able to achieve true "no map required for national highway NOA".

"With our current resources, it's mainly about accumulating data and grinding away the time. Achieving full high-speed operation shouldn't be a problem."

Wu Xinzhou said.

The highway conditions were quite simple for their team; they encountered almost no bottlenecks.

The training only requires different ramp scenarios and some unexpected scenario data that are difficult to collect.

For example, road obstacle recognition, such as road collapses, sudden stops of vehicles ahead, and other unexpected scenarios.

It's relatively rare on highways, but you can't say it doesn't exist.

In the event of such an incident, the vehicle-side system needs to make the correct judgment and identification.

They even have to consider situations where drivers mistakenly press the accelerator instead of the brake while following another car on the highway.

This falls under the scope of proactive AEB.

……

"There's also urban area data. I had Li Ligeng and his team collect it, but I have to say that the road conditions in China are much more complex than in Silicon Valley and Europe. The probability of achieving city-wide NoA in this generation is not high."

It's not just about using cameras to implement traffic lights and lane markings; it also involves dealing with frequent traffic jams and lane cutting, electric bikes suddenly appearing out of nowhere, tricycles and elderly people driving recklessly, and children and pedestrians—all of these need to be taken into account…

Some people really treat the highway like their living room. We didn't encounter these kinds of situations during our year-long training abroad, but they did provide a lot of valuable data for AEB training.

Wu Xinzhou thought for a moment, then gave a rare, bitter smile, as if he had a headache.

Xu Yi heard it.

"Old man's electric vehicle that does not follow the rules" +1.

He objectively stated, "This is not your fault. Current road regulations and driving etiquette in China are contributing factors and need to be improved."

"Currently, the computing power and transmission bandwidth of the vehicle are limited. The perception task of the map-less NoA in urban areas is too heavy. It requires real-time identification of 20+ categories of targets, which requires extremely high recognition accuracy. In addition, dynamic path planning and real-time adjustment of decisions require computing power that is five or six times that of high-precision map navigation."

However, we have also reserved upgrade interfaces, so once the high-speed NOA is fully operational, the city model training platform can be upgraded to one with higher computing power.

Although the highway NOA function has only just been launched, Wu Xinzhou is already thinking about laying the groundwork for urban intelligent driving functions.

A lot of thought was put into the internal planning.

“Developing autonomous driving and training algorithm models is a great thing, but the things you’re considering right now are hard for ordinary users to understand,” Xu Yi said after a moment of silence.

Users don't care what you've considered.

Users only care about their own user experience.

"Yes, Mr. Xu, what you said earlier inspired me. We've focused our efforts on optimizing the parking function. Now, it not only achieves high-precision automatic parking, but also one-click exit, parking after leaving the car, and even parking lot memory functions!"

Wu Xinzhou is not the kind of person who doesn't know how to be flexible.

The opening of the NOA highway is definitely an exciting development!
The first automaker in the country to launch highway navigation assistance.

This title allows for so much speculation!

This also demonstrates just how successful Xingchen Automotive's transformation towards intelligence and automation will be.

But in high-speed scenarios.

After all, many car owners only drive their cars a few times a year.

Moreover, the NoA navigation system on the highway is not available on all sections. Some people think it's amazing, while others think it's not very meaningful.

So--

Are there any more exciting features?
some!

Brother, I have some!
They've even developed three of these amazing features in-house!
Wu Xinzhou took out a car key, pointed it at the test vehicle parked in the garage a few meters away, and pressed and held the start button.

Drip, drip, drip…

Two seconds later, the hazard lights on the test vehicle came on, and with a "tick" sound, it automatically shifted into gear and started moving, slowly driving straight out of the parking space.

In less than two seconds, with no one around, they drove straight out of the parking space.

"This is a 'straight in and straight out' vehicle summoning function."

Wu Xinzhou pointed to the vehicle, "In rainy weather or when the parking space next to us is too narrow, we can use our car key to press and hold the start button to start the car directly, allowing it to drive out of the parking space unattended at low speed."

However, this was only the first step in his demonstration.

The "vehicle summoning" function can be traced back to 13 when BYD released the Surui model, which already had this function.

The logic of straight-in and straight-out is not complicated. At the time, it was the first car company in the world to mass-produce a car with a low-speed remote-controlled "straight-in and straight-out" function.

He opened the car door and deliberately moved it a little to the side. The car's large screen immediately displayed a model of the nearby parking spaces.

Tap the parking space model lightly with your finger.

The mechanical voice of the starry sky sounded: "Parking space detected, please select automatic parking mode."

Two options pop up on the screen:

—Moor immediately.

— Park the car.

...

Wu Xinzhou clicked the "Leave the car and park" button, and then got out of the car.

Immediately afterwards, the driverless test vehicle, with its hazard lights on, deftly reversed and smoothly backed into the parking space.

"It's not just simple parking spaces; we can achieve parking away from the vehicle. We also handle complex scenarios and narrow, irregularly shaped parking spaces. We've developed different solutions for each type of space, striving to ensure that machines can also back into the garage, just as people can back in."

Wu Xinzhou was very confident.

He then drove the test vehicle a short distance away.

To increase the difficulty, several more bollards were placed to simulate narrow and obstructed parking spaces.

The "parking off-vehicle" function was demonstrated once again.

This time, the test vehicle, without human intervention, turned the steering wheel three times and still managed to navigate around the cones and come to a steady stop among them.

He was quite proud.

"In this situation, you can not only park the car away from the parking space, but also let the car leave the parking space automatically without getting in the car."

Many people don't trust the navigation feature, or it was indeed incomplete when it was first launched.

So, automatic parking and automatic parking after leaving the vehicle, as long as the accuracy is high, are things that greatly improve the user experience.

They can also boast about their intelligence.

Xu Yi's eyes lit up slightly.

"not bad."

Its obstacle parking ability is probably twice as good as some other cars' reversing ability.

He was seized by a playful impulse.

I couldn't resist setting up a few more bollards. This time, I not only placed them on both sides of the parking space, but also restricted the space in front of the parking space.

After driving the car out.

To be honest, Xu Yiping would have had a hard time pouring it in from that angle himself, and he wasn't entirely confident.

"Try it again and see if you can achieve unmanned parking this way."

"President Xu, this..." Wu Xinzhou was dumbfounded when he saw it!

"Fine, I'll give it a try."

He sighed, got into his car, locked in a parking spot, and then got out.

The test vehicle clearly hesitated after recognizing the parking space.

Within the confined space, it kept crawling back and forth while the steering wheel was constantly turned.

The sound of tires rubbing against the ground continued.

This picture.

If someone saw this, they would be shocked!
The car behaved like a slightly impatient driver, constantly turning the steering wheel, but upon closer inspection, there was no one inside.

Because the space in front of the garage was completely blocked off by Xu Yi to the extreme.

After turning the steering wheel seven or eight times, the rear of the car was halfway in, and then it got stuck with its hazard lights on.

A prompt appeared on the car's infotainment screen:

"Unmanned parking failed. Please take manual control."

...

Xu Yi walked over and took a closer look. "It looks like it's almost done, but the collision boundary between the front of the car and the cone seems a bit conservative in your design, which is causing it to get stuck."

Wu Xinzhou also scanned the distance between the vehicle and the cones: "Yes, if a person uses the naked eye to identify, it is possible to get within 15 centimeters to reverse and adjust, while for general parking, 20 centimeters is the limit of safe distance."

Both cameras and millimeter-wave radars have blind spot cutoffs, and this distance is left to prevent collisions in the blind spot.

Can it be improved?

Xu Yi asked.

If we can even reverse into this extremely difficult and narrow parking space created by human intervention, or parallel park, then at least 98% of parking spaces, and perhaps even more than 99% in the future, will be able to be parked perfectly.

What kind of concept is that?
A garage that a person can't even back into, but given enough space, a vehicle can back into it to the limit.

That's true technological progress!

Instead of putting on a show with an automatic reverse parking function, only to have it poorly designed, automatically exiting at the slightest difficulty, or crashing straight into it.

That's like paying for a useless feature.

"can."

Wu Xinzhou nodded.

At this stage, if we don't focus on urban intelligent driving and concentrate our efforts on dealing with these extreme, irregularly shaped parking spaces, he feels there shouldn't be any problems.

He believes that achieving autonomous driving in urban areas is the real big goal, but there are many factors affecting it, including computing power, data, algorithm models, and external factors, so it cannot be achieved overnight.

This parking function.

If it weren't for Mr. Xu's request, he really wouldn't have thought of doing a deep optimization and developing based on it.

Now that Mr. Xu has said so, he must have his reasons.

Those of you with technical skills can handle the problem.

"In addition, we also have a parking lot memory function and an AEB function that works across the entire speed range of 4-135km/h!"

Wu Xinzhou called another car and demonstrated it again.

This time, the car was outside the test track.

The so-called parking lot memory parking function is to memorize a fixed area of ​​a parking lot.

After arriving at the community parking lot, the car owner can click on the memory parking function, and the vehicle will automatically recognize and activate the function, following the memory.

This feature is suitable for indoor parking lots with fixed parking spaces.

The demonstration was successful on the first try.

Wu Xinzhou demonstrated the "full-speed domain AEB function" several times using a dummy and foam obstacles in the test field.

For vehicles currently on the market, the conventional AEB (Autonomous Emergency Braking) solutions developed by suppliers and various companies can only support speeds up to 4-80 km/h.

The range exceeds 80km/h.

This function is not working.

The so-called "automatic risk avoidance" and "automatic braking" advertised by car manufacturers no longer exist.

of course.

This feature isn't about who has more support for a wider range of applications winning.

The most crucial factor is still "recognition accuracy"!
If it's triggered accidentally, for example on the highway, or when you're driving normally and suddenly brake to a stop, then it's an accident.

so far.

Apart from Volvo and a few other automakers that emphasize safety as a selling point, almost no other automakers put effort into this aspect.

Even if you do it well, the returns won't be significant.

Doing it poorly might result in being criticized or even causing an accident.

It's better to just use a supplier's solution and give it a sloppy pass.

"We have fully integrated the AEB function into intelligent driving perception, unlike most models on the market that only support forward collision avoidance."

But our vehicle can also support side collision avoidance and rear braking...

It can recognize and stop stationary vehicles at speeds up to 90 km/h, and even if a vehicle suddenly appears from behind a vehicle, it can still avoid collisions at speeds up to 60 km/h.

The test success rate is over 95%.

This feature is not easy to demonstrate.

Wu Xinzhou simply gave a brief explanation.

They've actually put more effort into "AEB safety" than into driverless parking and parking lot memory functions.

After all, it involves "safety".

Safety is the bottom line.

Safety is the greatest luxury!

Even if we take a step back and consider the worst-case scenario, sometimes a single serious safety incident can have a more devastating impact than a product quality scandal.

Especially the domestic public opinion environment.

If intelligent technology and smart driving are used for advertising, but this results in serious accidents, then...

No matter how great the product is, it can still lead to a major setback in reputation and brand image.

……

"Two years ago, Volvo made a fool of itself because of this. They were testing AEB in Malaysia, traveling at less than 20 mph, and ended up getting hit and thrown into the air..."

Wu Xinzhou chuckled and said, "But I can guarantee that our cars can stop almost 100% of the time in this kind of test scenario at speeds up to 50 km/h. However, physical testing is not advisable. We hope that the efforts we make internally will be used to encourage car owners to spontaneously spread the word when they encounter any emergency situations in the future."

This joke has been mocked online for two years now.

Reference – [Volvo CEO was accidentally knocked off his vehicle while testing AEB].

According to subsequent media reports, the person who was hit was not a president at all, but a Malaysian dealer who sold a low-spec XC60 as a high-spec model. It was not equipped with a pedestrian safety detection system and could not possibly have triggered AEB.

However, because of this matter...

Volvo subsequently made AEB standard on all its lower-spec models as well.

The unexpected result of that "commendable courage" warrior's contribution to industry progress left onlookers both amused and bewildered.

　The initial inventory of mobile phone battery cells was calculated incorrectly in the previous chapter; it has been corrected.

　　
　
(End of this chapter)