Jia Haonan from Wafei Temple
量子位 | 公众号 QbitAI
The world's first intelligent heavy truck with the ability to take off your hands and eyes is here!
有啥用?
In the past, assisted driving made truck drivers become "safety officers", but now they are manned and machine in turn, and they are properly second drivers.
It means that the driver's work is easier, and the length of the route that a single person can carry has also increased significantly compared with the past, which means lower costs for cargo owners and higher income for drivers.
It is the eternal pain point and core concern of the logistics and freight industry.
The products and technologies behind the "hands-off and eye-off" smart heavy trucks come from Qianhang Technology, an autonomous driving company established less than 3 years ago.
Who is Qianhang, why was it the first to achieve "hands-off and eyes"? and most importantly, is it safe? Can it run commercially?
What volume did Qianhang Technology submit?
Hands-off and eye-off is generally considered to be a landmark function of high-end intelligent driving in the transition stage from "human-machine co-driving" to "unmanned driving", "human-machine rotation driving".
However, the requirements for the system's capabilities are even higher: before the safety officer had to keep an eye on the road conditions and be ready to take over at any time, most of the time, they can take their attention off the road conditions and the system can still operate safely and smoothly.
There are individual attempts in the field of passenger cars, but the landing is still far away. However, who would have thought that such an intelligent driving function would be the first to be available on heavy trucks.
This is the first delivery of Qianhang Technology - AutraOne, the official full name is "intelligent driving transportation tool", for the trunk logistics and transportation market, and the delivery will begin in the second quarter of this year.
For trunk logistics, first of all, it means that AutraOne is a heavy-duty truck, based on Dongfeng Liuzhou's Chenglong H7 heavy-duty truck tractor, customized with 130 square meters of low-wind resistance van trailer:
The point is that AutraOne is not modified on the H7 finished vehicle, but based on the H7, after in-depth cooperation between Qianhang and Dongfeng Liuzhou, from the structural restructuring, sensor layout, power chain optimization, intelligent cockpit configuration, underlying electronic and electrical architecture upgrade, software algorithm development and other aspects of the development of the front-loading mass production intelligent truck.
This also makes AutraOne's intelligence more stable, consistent, reliable, and more compliant and economical.
The AutraOne is equipped with six lidars, seven high-definition cameras, three millimeter-wave radars and six ultrasonic radars. Compared with the L2 assisted driving system and L2+ high-speed pilot assisted driving system on the market, the number and types of sensors are much more.
At the same time, the underlying computing platform, the self-developed AutraSpark Max has 1824K DMIPS general computing power (CPU) and 1800 TOPS neural network inference capability (NPU).
Equivalent to 8 Orin X computing power, this underlying computing platform is currently the largest in both passenger cars and commercial vehicles.
Interestingly, Nvidia has just officially announced the strongest car core Thor at CES in 2025, and there are mass-produced passenger cars in China and 8 OrinX computing power are planned for the 2000T level, I don't know if this AutraSpark Max is a replacement product before the transition to the Thor platform in 2025, and it will compete with China's head mass-produced passenger cars.
In addition to the large computing power, AutraSpark Max is designed with a three-layer structure with reference to the aircraft standard, which can not only work independently, redundant backup for each other, but also divide labor and cooperate to avoid homologous failure through different systems and chips.
In terms of algorithms, the self-developed AutraPilot includes AutraOS, a communication and scheduling system that integrates Rust and C++ languages, a full-scene prediction algorithm using Transformer architecture, a planning algorithm that integrates spatio-temporal union with full analytical solutions, and a control algorithm that takes into account safety, smoothness and fuel saving.
In terms of specific functions, AutraOne has the ability to independently complete all driving behaviors on the highway, including going straight on the main road and ramps, following, avoiding, overtaking, merging, merging, merging, exporting and other operations, with a maximum speed of 100km/h, which is also the maximum speed limit for heavy trucks on the mainland.
According to the official introduction of Qianhang, AutraPilot currently has the technical ability of "toll station - toll station", users take off their hands and eyes, and theoretically do not need to keep an eye on the road conditions and system at all times.
These are the first submissions of Qianhang Technology's 2 and a half years of entrepreneurship.
The answers given to the smart heavy truck are obviously different from others, and even a little "abnormal". For example, the cost of intelligent driving hardware is by no means the lowest price choice, which seems to violate the consensus of "cost first" in the logistics industry.
Why did Qianhang choose such a product route, and can the commercialization logic run through?
Smart heavy trucks, why do you want to "take off your hands and eyes"?
The most intuitive factor, because to realize the intelligent driving function of "taking off the hands and eyes", at the perception level, it is necessary to fill the capacity and make redundancy.
So the question becomes: why do smart heavy trucks have the function of "taking off their hands and eyes"?
Of course you can, but it's not enough.
The average length of China's first-class trunk line (i.e., the inter-provincial access road planned and managed by the state) is 1,200 kilometers, and the one-way driving time is about 17 hours. Most of the heavy-duty truck semi-trailers and trunk logistics with relatively stable supply/demand are concentrated on such lines.
17 hours away, many goods still have timeliness requirements, traditional logistics companies usually use the form of 2 drivers in shifts. In the past two years, the smart truck track has put forward the strategy of "two driving to single driving", reducing the burden on drivers through intelligent driving functions, and trying to save costs by reducing one driver.
But the contradiction is also obvious, even if you are only a safety officer, you will be staring at the whole process for more than ten hours, which will inevitably exceed the physical limit and bring new potential safety hazards.
Such a dual-drive to a single drive, the line that can be carried is limited, generally below seven or eight hundred kilometers, and most commercial vehicle auxiliary driving can not reach the speed requirements of 100km/h, and the list with high timeliness requirements cannot be met.
Therefore, the necessity of "taking off the eyes and hands" is reflected, only in this way can a driver easily run the long-distance journey, and a car can really save at least 15-180,000 yuan per year.
Qianhang Technology calls such a function man-machine rotation driving.
In terms of specific safety measures, the pursuit of Qianhang Technology at this stage is not to let the intelligent driving system pass through all extremely complex scenes autonomously, but to ensure the successful detection of all obstacles and abnormalities on the road as much as possible.
Once detected, in the case that all levels of system functions cannot be circumvented or bypassed, there is one last redundancy: pull over, or even brake directly, and manually take over after parking.
"Rotation" does not mean co-driving, but normal road conditions do not need to be worried, once there is an abnormality, it is guaranteed to stop in advance to avoid major accident losses.
That's why AutraOne's sensor and computing power configuration has reached an unprecedented height.
For example, the forward wavelength 1550nm lidar has a stable detection distance of 300m and an obstacle limit detection distance of 500m, in order to meet the safety requirements of heavy trucks with longer braking distances.
In addition, the introduction of 70-degree ultra-wide-angle blind lidar can shorten the blind area of the lidar from 13m to 30cm, and cooperate with the secondary enhancement of ultrasonic radar to achieve three redundant and stable detection of people/objects in front of 20cm.
The medium and long-range lidar at the side and rear can provide a safe distance of 180m for merging. In addition, the redundancy of mmWave radar supports autonomous and safe parking capabilities in low-visibility weather, when the sensors are dirty and obscured.
This is also the core competitiveness of the commercialization of Qianhang Technology:
The long-distance trunk line really reduces a driver, and uses "braking" as a safety guarantee.
There are hard 15-180,000 labor cost reductions a year, and with the self-developed fuel-saving algorithm (about 5% fuel saving in actual measurement), according to the calculation of Qianhang Technology, the cost can catch up with traditional trucks in the first full year of operation without a heavy truck, and every year after that, more benefits can be created by reducing costs.
The increase in the cost of "luxury" intelligent driving kits is actually leveled by the understanding of logistics scenarios and models.
In addition to the above-mentioned fixed-route orders that change from two-drive to single-drive, it can also allow single drivers to undertake scattered orders on longer routes (such as 2000km from Beijing to Dongguan) and ensure better timeliness.
In addition, compared with human drivers, intelligent driving algorithms are more stable and cautious, which can bring further benefits such as reduced cargo damage and insurance costs.
One more thing to mention here is that the country has just opened the L3 pilot license, and Qianhang Technology believes that this is a major benefit to the business.
Because of the release of L3 regulations, the existing technical system of Qianhang can quickly make trunk freight intelligent heavy trucks covering more than 1,200 kilometers, which can reproduce the advantages of "reducing drivers and reducing costs" in a larger market range.
To summarize the commercialization of Qianhang Technology, the "three-board axe":
Front-loading mass production achieves a balance between cost and performance, true dual-drive to single-drive, and fuel-saving algorithms. The core of it is to change from two drivers to single driving.
In fact, in the smart truck track, this set of commercialization logic is not the first one, and many players have put forward the same idea before. However, the question of the feasibility of "double driving to single driving" has not stopped, mainly because it is difficult to solve the problem of breaking through the physical limit of safety personnel for long mileage.
Senkake is the first player to solve this problem with a clear technical solution.
A new stage of intelligent heavy truck racing
Qianhang Technology is a very new face in the smart truck track, established in 2021.
The core team includes high-achieving students from Princeton, Carnegie Mellon CMU, Peking University and other famous universities, including long-term tracking and investment in the field of smart cars and autonomous driving, as well as technical experts who have been in charge of truck projects in leading autonomous driving companies.
In addition, there is Liao Ruoxue, who is known as the "technology ceiling" in the industry, and is said to be the big guy who puts Robin Li's name on the Baidu patent wall.
In the past few years since the establishment of Qianhang Technology, it has also been the three years when the smart truck track has shaken and shuffled and moved towards a turning point.
During this period, start-ups are moving towards the goal of mass production, but they either do not cooperate well with OEMs, lack of engineering experience, or encounter policy resistance, etc., in short, the long-term commercialization is not smooth, which directly triggers their own crises of varying degrees.
Since last year, the industry has gradually formed a consensus on the implementation of technology and the construction of a closed business loop.
This consistency is also evident in Qianhang Technology.
In less than 3 years of mass production and delivery, Qianhang reflects the necessary capabilities of smart trucks to get the admission ticket: software and hardware research and development, front-loading mass production, and commercial closed-loop.
In fact, whether you are entering the game to build a car or cooperating with an OEM, these capabilities are necessary for smart truck players.
However, Qianhang Technology has shown its "uniqueness".
In terms of technical solutions, the cost of intelligent driving is full, and the "excessive" technical investment is actually the lower limit of safety and the upper limit of experience of smart heavy trucks, and at the same time it is very high, at least at present, no other player has reached such a height.
The reason for choosing such a solution is that Qianhang Technology has invested a lot of energy and resources to investigate the logistics and freight track, subdivide the business scenarios in units of 100 kilometers, find out the neglected pain point needs, and then put forward the corresponding plan, answer the questions that no one in the industry has answered before, and find the commercialization path that no one has walked.
In addition to autonomous driving companies, Qianhang has more genes of "logistics companies".
Qianhang Technology released the first model to start commercialization, which is actually a sign of the new stage of competition of smart trucks.
The development of intelligent driving software and hardware architecture tends to mature, and the focus of competition has shifted to landing scenarios, functional requirements, cost calculations, etc.
In the face of different logistics and freight scenarios, they may have their own best intelligent driving solutions, in fact, all players in the industry have the opportunity to quickly overtake.
Of course, this requires players to have an in-depth understanding of details such as logistics business and link costs, as well as a high degree of sensitivity to technology applications.
Under the new trend, Qianhang Technology is worth paying attention to.
— END —
QbitAI · Headline number signed
Follow us and be the first to know about cutting-edge technology trends