Zhuoyu Releases Commercial Vehicle Intelligent Driving Solution "Panxing": PDS 4 Already in Mass Production, Payback in One Year

On the evening of October 29, intelligent driving solution provider Zhuoyu Technology officially released its commercial vehicle intelligent assisted driving solution "Zhuoyu Panxing," and announced business progress in commercial heavy trucks, buses, light trucks, ROBOVAN and other areas.

This Zhuoyu Panxing is divided into two versions, PDS 4 and PDS 5.

PDS 4 corresponds to the highly perceptive large model version 4.0 and above, mainly supporting highway NOA and targeting heavy truck trunk-line logistics. This version has already achieved mass production.

PDS 5 corresponds to Zhuoyu's next-generation native multimodal foundation model. In addition to highway NOA, it will also support urban NOA, mainly targeting commercial vehicle models such as buses, light trucks and VANs that need to drive extensively on urban roads.

Why did Zhuoyu choose this time to enter commercial vehicles? Zhuoyu Technology CEO Shen Shaojie said that the current penetration rate of intelligent driving in domestic commercial vehicles is less than 1%, while the penetration rate of L2+ in passenger vehicles has already exceeded 30%. He expects that in the next two to three years, intelligent driving in commercial vehicles will enter a faster growth phase.

Over the past few years, intelligent driving in passenger vehicles has already gone through a round of rapid popularization. The costs of cameras, millimeter-wave radar, domain controllers and high-compute chips have continued to decline, and technologies such as end-to-end and large models have also entered mass-produced models.

For suppliers that already have passenger vehicle customers and mass-production data, migrating existing technology to commercial vehicles costs less than developing a completely new system.

In Shen Shaojie's view, an important advantage of Zhuoyu Technology is data.

The cost of collecting commercial vehicle data is significantly higher than that of passenger vehicles. In addition to the vehicles themselves, fuel costs and tolls are also higher. In the early stage of the business, before the commercial vehicle industry's own data system has been established, if data were collected entirely anew using commercial vehicles, the investment would be very high. Zhuoyu Technology already has passenger vehicle data and a model foundation, which can lower the upfront cost of entering the commercial vehicle market.

But intelligent driving for commercial vehicles is not low in difficulty.

Zhuoyu Technology Vice President of Commercial Vehicle Business Liu Sikang said that intelligent driving for commercial heavy trucks has at least five difficulties: little model data, high requirements for perception distance, low safety fault tolerance, great difficulty in vehicle control, and highly sensitive costs.

A fully loaded heavy truck can weigh around 40 tons. When driving at high speed, the body is long, blind spots are large, and braking distances are longer. In addition, the center of gravity and dynamic state of a heavy truck differ greatly between empty and fully loaded conditions. For a semi-trailer of about 17 meters, issues such as inner wheel difference and trailer sway also need to be handled during turning and lane changes.

Zhuoyu did not develop a completely independent intelligent driving system anew for commercial vehicles, but directly migrated the large model already trained for passenger vehicles.

Liu Sikang summarized its technical route in three phrases: unified foundation, scenario enhancement, and engineering adaptation.

Zhuoyu believes that although passenger vehicles and commercial vehicles are different models, they mainly run on paved roads, and basic capabilities such as road structure recognition, traffic participant recognition and risk judgment have strong commonalities.

At present, Zhuoyu's passenger vehicle intelligent driving system has accumulated more than 3 billion kilometers of data. On this basis, Zhuoyu added about 400 million frames of heavy truck-specific data to conduct targeted training for heavy truck trunk-line transportation scenarios. It is reported that this solution has achieved mass production of an end-to-end large model for commercial heavy trucks.

The second part is engineering adaptation. Because heavy trucks have longer braking distances, they also have higher requirements for long-distance perception.

The Panxing solution is equipped with Zhuoyu's "Jimu 2.0" system. Compared with installing lidar outside the vehicle, Jimu 2.0 is placed inside the cab, which can reduce the impact of rain, snow, mud and gravel on the sensors.

According to data released by Zhuoyu, this system can achieve a detection range of up to 500 meters in highway scenarios and can identify low-reflection, small-sized obstacles such as distant tires and cardboard boxes.

Vehicle control also needs to be re-adapted. Zhuoyu uses vision to recognize trailer posture and estimate the angle between the tractor and the trailer, while combining real-time load information to adjust vehicle steering, braking and power control.

For uphill and downhill sections, continuous curves and complex road surfaces, the system also needs to coordinate different actuators such as the engine, retarder, auxiliary brake and foot brake.

However, compared with passenger vehicle users, if an assisted driving system cannot reduce fuel consumption, reduce accidents or lower driver workload, it will be difficult for it to become a long-term purchased configuration for commercial vehicle customers. Therefore, Zhuoyu also added a fuel-saving algorithm to Panxing.

According to Zhuoyu, this algorithm is trained using reinforcement learning and calculates the timing for releasing the accelerator, applying the accelerator, and using the retarder and braking system based on road slope, curves, the vehicle powertrain and surrounding traffic conditions.

Over the past two months, Zhuoyu has tracked more than 200 mass-produced heavy trucks, with cumulative operating mileage exceeding 20 million kilometers. Data released by Zhuoyu shows that after being equipped with this algorithm, the vehicle fuel-saving rate reached 4.5%-7.5%.

According to Zhuoyu's calculations, based on driving 800 kilometers per day, a heavy truck can save about 80 yuan in fuel costs per day, or about 20,000 to 30,000 yuan per year.

If the reduction in driver fatigue, safety accidents and insurance costs is also taken into account, Zhuoyu expects that an intelligent assisted driving system can pay for itself in about one year.

At present, Zhuoyu has reached cooperation with six major domestic heavy truck companies: XCMG Automobile, Shaanxi Automobile Group, Sinotruk, FAW Jiefang, Foton Motor and Dongfeng Commercial Vehicle. According to Zhuoyu's statistics, these six companies together account for more than 93% of the domestic commercial heavy truck market.

Among them, Shaanxi Automobile is the first mass-production customer. In June this year, the first Shaanxi Automobile Delong X5000 equipped with Zhuoyu Panxing PDS 4 was officially mass-produced, and hundreds of vehicles are currently operating on Yunda Express trunk lines.

In September this year, XCMG released the Hanjing new energy heavy truck, and models equipped with Zhuoyu Panxing PDS 4 are expected to be delivered by the end of this year. Sinotruk's high-end heavy truck Sitrak also plans to roll off the mass-production line by the end of this year.

In overseas markets, Zhuoyu has already advanced overseas cooperation with customers such as Sinotruk, Foton Motor and Yutong Bus. Among them, the European project in cooperation with Sinotruk has completed part of the compliance preparations and is expected to land next year.

Zhuoyu has completed the mass production of the first batch of models. Next, whether this business can continue to expand will depend on whether indicators such as fuel savings, reliability and payback period can be verified in the long-term operations of more fleets. Leiphone Leiphone