Hao Nan Lei Gang sent from the Vice Pilot Temple
Smart car reference | Official account AI4Auto
Bon friends, if you want to talk about the king of automatic driving technology, is it not a pony Zhixing?
Just last year, they showed a video of a completely driverless Robotaxi in the city of Guangzhou that had not been taken over for 1 hour, and the two founders in the back seat were too slippery to sit down - the strength of the technology, the confidence of the strength, people have nothing to say.
But unexpectedly, the volume king has recently rolled to new heights.
No, Pony Zhixing has just exposed a new daily routine:
Autonomous driving that does not take over for 6 consecutive hours at high speed in China.
And not only the length of the roll, this time the protagonist is also replaced by a truck - through the night to day, through the long tunnel, through heavy rain and fog weather, road encounters the road fault car turtle speed car and special-shaped trailer and other real scenes.
In the end of 6 hours, many technical details are worth savoring.
Analysis of the highlights of Xiaoma Zhika video technology
The autonomous driving demonstration of trunk logistics has also had many representative works in the industry before.
Including the road conditions in the United States, there is also the feat of dropping the safety officer completely in the past.
But this display of Xiaoma Zhika has three industries first:
First, the industry's longest continuous autonomous driving does not take over.
6 hours of time, passenger and freight, never before. The single MPI range is 405 km long – a total of 405 km during this 6-hour journey.
Second, technical implementation under complex Chinese road conditions.
Different road conditions and scenarios in different regions undoubtedly have a huge gap in the challenges of the landing of autonomous driving technology.
And this is also (almost) the first time that the landing of autonomous driving trunk logistics has been demonstrated in such a one-shot way.
The complexity of the road conditions, the changeable scenes, and the early and dark status quo on China's trunk line are also clearly recorded.
Third, as a company in the industry that promotes passenger autonomous driving and high-speed freight autonomous driving at the same time, Xiaoma Zhixing used this video to answer the once controversial question in the industry - is there an unbridgeable gap in the advancement of autonomous driving technology for passenger cars and trucks?
According to Xiaoma Zhixing, the mileage exceeded 405 kilometers in 6 hours, experienced 5 different highways, 13 tunnels, performed 4235 steering times, performed acceleration \ braking 1052 times, and tracked and identified 4806 dynamic traffic objects... Friends who have time can study it carefully.
We will combine the "lite version" and mention 9 technical highlights worth paying attention to.
1. From dark to dawn
In this video explanation, it is clear that the departure time is 5 am.
It was not yet light, and the ground was wet because of the rain – making it harder to drive the truck.
In the black sky mode, the core sensor capabilities such as cameras will theoretically be discounted, so just set off, Xiaoma Zhika quietly showed a wave of technical confidence without saying a word.
2. The large curvature ramp merges into the main road
After driving in from the toll booth, the PonyChika started the automatic driving mode.
But new challenges also follow, requiring the completion of the convergence of the ramp to the main road, and the integration of other ramps into the main road, mainly depending on the busyness of the traffic, which needs to be decisively integrated at the right time.
However, the main thing about the Pony Zhika this time is the control ability, because the ramp curvature is larger - it is more "curved", the ramp is narrow, and the truck is a multi-rigid body structure, the front and the trailer are separated, and the control requirements are higher.
From the final performance, the control is very precise, whether it is the front of the tractor or the body of the trailer, it is strictly controlled within the lane line.
3. Static fault car avoidance on high speed
That's right, high-speed, stationary, faulty cars.
These keywords put together have made many so-called "automatic driving" but only assisted driving ability of the system accident on the news hot search.
This time, The Little Ma Zhika also encountered, and a faulty car with a double flash appeared in the lane ahead.
But just about a hundred meters away, the pony chika showed that the identification was completed and showed a wave of active lane change avoidance.
In fact, the faulty car was in the emergency lane, and the Xiaoma Zhika was driving in the rightmost lane at the time, but it should be for the reason of "eliminating any possible safety hazards", actively avoiding, overtaking, and finally returning to the rightmost lane again.
In addition, The little horse Zhika avoided the faulty car this time, in fact, there is a more difficult external challenge.
4, there is wind, rain and fog
When PonyChika first hit the road, it was a rainy day, and there was clearly visible water on the road.
The more rain it gets, the more it can be seen through the gear of the wiper. And other vehicles passing by, all of them made a lot of splashes.
In addition, the whole outside is also a clear foggy environment.
This kind of weather with wind and rain and fog, the sensor perception ability, and the control of the vehicle itself on the slippery road, are more tested than the usual weather.
On the one hand, the performance of D.C. Chika illustrates its multi-redundant sensor solution , including lidar, millimeter-wave radar and cameras – with a unified, stable output.
On the other hand, it also shows that in the high-speed trunk logistics scenario, pure vision or a single sensor scheme may not be able to cope with such a scenario.
5. Construction area
Ma Zhika also passed by an area of construction ahead and took the initiative to change lanes to avoid it.
First change lanes to the left lane, slow down as you pass through the construction area, and then change lanes back to the right lane after passing.
And it is worth noting that because the construction prompts are mainly through signage, this wave of system operation, invisibly shows the recognition accuracy of camera vision.
6. 4 km long tunnel
Yes, 4 km long tunnel.
The biggest technical challenge of tunneling is signal loss.
In the solution of Xiaoma Zhika, there is a GNSS signal positioning system including GPS, but there will be signal loss under the tunnel in general.
Even human driving may be more difficult, requiring more refreshment and concentration, let alone AI drivers.
However, Xiaoma Zhika did not show "discomfort" in the entire long tunnel, driving smoothly, keeping the lane very stable, and strictly adhering to traffic rules - turning on the lights in the tunnel, reducing lane changes, strictly prohibiting solid line lane changes, and so on.
It is said that behind this is still the technical power of the deep integration of various sensors of Xiaoma Zhika.
The positioning system of autonomous trucks makes full use of laser point cloud matching, lane line matching, odometer inertial guidance and other technologies to achieve reliable vehicle positioning and allow trucks to cope with the challenge of weak signals in the tunnel.
7. Special-shaped vehicles
In the process, I also encountered a trailer on the road, which was still empty.
The biggest challenge for AI drivers is that they are "seen less", sometimes mistakenly or not.
For example, there was a system of mass-produced passenger cars before, which directly hit the trailer in front of it at high speed.
But in this scene, Xiao Ma Zhika did not suffer losses. The specific reasons will be discussed later.
8. Turtle speed car on the highway
There are many of the most annoying scenes on the highway, and the turtle speed car is definitely one of the first in line.
In the video of The Little Horse, a slow-moving van is shown...
If the AI driver is "honest", he may follow the car.
But the AI driver of Xiaoma Zhika is obviously very confident, very similar to the old human driver, "observing the road conditions" and then taking the decision to change lanes, overtake, and then change lanes to come back.
A wave must be both safe and efficient to operate.
9. Drive off the ramp
Getting on the ramp is challenging, and driving off the ramp is no easy either.
However, under the previous series of technical confidence demonstrations, the on-ramp seems to be "scaring anyone".
And Pony Zhika has long been "aware" of when and where to remit it.
In short, in the video of Xiaoma Zhika, at least the above 9 scenes and details can show the technical level and strength behind it.
There is no doubt that this virtual driver can rival the old driver of a human truck.
And from the perspective of connecting points into a line, although the technology research and development of Xiaoma Zhika has been exposed in 2019, the large-scale landing and promotion is actually in 2021.
Combined with these several time points and the level now displayed, the speed and level of system iteration can also be seen.
What does Pony Chi Heng want to prove?
At least three points.
First of all, the most intuitive proof is that the truck automatic driving technology of Xiaoma Zhixing is also very "old driver", and in truck freight, the technical level of Xiaoma's AI driver and virtual driver is also amazing.
At the beginning, we also said that the level of passenger car autonomous driving displayed by Xiaoma Zhixing last year brought the industry technology to a new height.
But the camera of this wave of trucks driving themselves has rolled up to new heights.
Sure enough, the only thing that surpassed the Scroll King was the Scroll King himself.
Secondly, Pony Chika provides a new argument for another hot topic in the industry.
Whether passenger autonomous driving and truck autonomous driving can be promoted at the same time and complement each other has been debated endlessly and inconclusively.
The square thinks yes. Including Waymo, Aurora and Xiaoma Zhixing, because they are all ai virtual drivers, they all hope to drive humans out of the driver's seat, why can't technology be reused?
Even because the road condition scenario of truck trunk logistics is relatively simpler, there is still the possibility and feasibility of dimensionality reduction between passenger and freight.
But the opposing side agrees that "I just want to be beautiful." The reason is that the unique challenges of high-speed trunk lines and trucks make this dimensionality reduction theoretically possible, but it lacks the feasibility of actual landing, and whether the data can be reused is another matter.
There are entrepreneurs who drive themselves by trucks and freight, and there have been well-known rhetorical questions: There are cars and trucks parked downstairs at the same time, there are many people who can drive away cars, how many can drive away trucks?
Therefore, there is no consensus in the industry as to whether cars and taxis can be promoted and mutually beneficial at the same time, and whether there is a gap in technology and landing.
However, through the display of The Pony Zhixing this time, it proved that the trolley and the big car can be pushed at the same time, and they can all advance very well.
Of course, the proof of the pony may only be its own strong technical strength and deep moat, which cannot give a final word to the positive and negative arguments in the industry.
Finally, whether it is the technical strength displayed by Xiaoma Zhika, or the recent business landing and promotion, it points to a big topic:
The truck business of Xiaoma Zhixing has not decreased in speed and is still moving forward at a high speed.
It is a response to the question with action.
Also in the second half of last year, Xiaoma Zhixing's truck business underwent a wave of internal adjustments, when the research and development of small cars and large vehicles was unified in the way of technology middle-office integration.
At one point, it was thought that Pony's two-legged strategy had been shaken – and repeated external explanations had not helped.
In the end, Pony Zhixing used the method that Xiaoma Zhixing was best at - through the demonstration of technical strength, it proved the steadfastness and advancement of truck freight, and it was still very fast.
In addition, another background is that Xiaoma Zhika has obtained the license for the first public road test of L4 self-driving trucks in China, and has technical proof at the regulatory level.
As well as commercial landing, Xiaoma Zhixing also established a joint venture with Sinotrans, the largest port logistics freighter in China, to establish a smart logistics subsidiary, Qingxiao Logistics, which has begun formal operation on April 1, opening a commercial large-scale landing journey.
OK, the above is the latest progress of Pony Zhika and Xiaoma Zhixing, and finally attached to the full version of the 6-hour video, a little high-energy, it is easy to watch in one breath, try not to open the viewing at a time that may cause insomnia.
Just sauce ~
—Ends—