Article | AI Relative Theory

L4 autonomous vehicles appear to have reached a pivotal juncture for widespread deployment.

The excitement surrounding the year 2025 remains palpable, and the Cainiao Group has stoked the flames with the official launch of its new L4 autonomous vehicle for public roads—the Cainiao Autonomous Vehicle GT Pro. Equipped with Cainiao's proprietary L4 autonomous driving technology, this vehicle independently plans routes, avoids obstacles, and safely navigates complex urban road environments, providing express delivery services to terminal stations.

If L4 autonomous driving technology is applied on a large scale, autonomous vehicles will significantly enhance logistics efficiency, reduce costs, and hold immense potential to transform the current last-mile logistics and distribution paradigm.

Concurrently, within the industry, vendors such as Jiushi Intelligence and Neolix have also introduced L4 autonomous vehicle products. Unlike the evolution of intelligent driving in passenger vehicles, the L4 upgrade for autonomous vehicles may arrive sooner and more comprehensively in 2025.

Autonomous Vehicles Fully Embracing L4 Autonomous Driving

As is widely known, L4 represents a relatively advanced level within the autonomous driving technology stack, relying on high-precision sensors (such as LiDAR, cameras, millimeter-wave radars, etc.) to accurately perceive the surrounding environment. Through complex algorithms, it processes and analyzes collected data to enable autonomous navigation and driving decisions for vehicles.

In layman's terms, vehicles equipped with L4 autonomous driving technology possess a high degree of autonomous driving capability, capable of autonomously completing driving tasks without continuous driver intervention.

So, are L4 autonomous vehicles equivalent to L4 passenger vehicles?

The answer is not entirely affirmative. The realization of L4 autonomous vehicles hinges on L4 autonomous driving technology. The application scenarios of L4 autonomous driving are broader, encompassing various types of vehicles in different fields, including passenger vehicles and commercial vehicles, as well as highway autonomous driving and urban road assisted driving. However, L4 autonomous vehicles primarily focus on specific operational scenarios such as logistics and distribution, environmental sanitation, and park patrols.

In essence, L4 autonomous vehicles exhibit a strong application-specific attribute and are a holistic product of technology and scenarios. Moreover, being fully unmanned, L4 autonomous vehicles face unique regulatory challenges and require a novel regulatory framework to govern their operations, encompassing vehicle access standards, stringent restrictions on operating areas, and mechanisms for accident responsibility identification. The formulation of these regulations must consider how to ensure public safety and social order in the absence of drivers.

For the industry, with the introduction of Cainiao GT Pro, autonomous driving and unmanned delivery may accelerate at a faster pace in last-mile logistics and distribution scenarios, further propelling the transformation of autonomous vehicles in this sector. At least in 2025, more autonomous vehicle manufacturers and last-mile logistics and distribution companies will need to focus on L4 applications, signifying the dawn of accelerated industry development.

Why the Emphasis on 'Public Roads'?

In fact, whether autonomous vehicles can successfully complete delivery tasks on public roads poses a significant challenge to their development. When autonomous vehicles enter public roads, it signifies a high level of maturity in vehicle driving, possessing the ability to interact with pedestrians, moving vehicles, traffic conditions, and other real-world environments.

Prior to this, autonomous vehicles underwent extensive testing and operations on test roads and semi-open roads. As of 2024, Cainiao autonomous vehicles have operated on semi-open roads in universities across more than 20 provinces nationwide, accumulating a total mileage exceeding 5 million kilometers and completing over 40 million order deliveries, amassing rich practical experience.

Building on this rich practical experience, we may witness more deployments of autonomous vehicles on public roads in 2025. Previously, during the 2024 Double 11 shopping festival, a comprehensive express delivery outlet in Yuhang District, Hangzhou, entrusted more than half of its packages to autonomous vehicles for delivery.

The continuous enhancement of autonomous vehicles' perception and autonomous driving capabilities undoubtedly unveils numerous possibilities for the industry's next phase of development. Based on currently publicly available solutions and products within the industry, we can also observe that the "eyes," "brains," "ears," and other senses of autonomous vehicles are being fortified like never before.

In terms of "eyes," autonomous vehicle manufacturers are bolstering the vehicle's recognition and perception capabilities through advanced multi-sensor fusion technology, the introduction of rain and snow filtering algorithms, sensor self-cleaning systems, and the integration of generative AI technology.

Among them, Jiushi Intelligence employs a perception solution comprising 14 cameras and 4 automotive-grade solid-state LiDARs, whereas Cainiao utilizes a multi-sensor fusion perception solution featuring 1 LiDAR and 11 high-definition cameras. These comprehensive solutions facilitate improved environmental perception capabilities, effectively identifying changes in the vehicle's surroundings, and even enhancing environmental perception in extreme weather conditions through algorithms and AIGC technology, which will substantially enhance the "vision" of autonomous vehicles.

Considering that public roads are far more intricate than we imagine, it is imperative to comprehensively consider traffic rules, pedestrian reactions, other vehicles' movements, and animal appearances, and make accurate and prompt decision responses. Autonomous vehicle manufacturers continue to upgrade and optimize the vehicle's "brain" to endow it with more sensitive and powerful thinking and reaction capabilities.

Among them, Cainiao autonomous vehicles adopt self-developed L4 autonomous driving technology and integrate artificial intelligence technology from Alibaba's DAMO Academy, enabling the vehicles to possess human-like cognitive intelligence. The brain's emergency response speed is seven times faster than that of humans, capable of discerning the intentions of over 100 pedestrians and vehicles in just 0.01 seconds, effortlessly handling complex road conditions and selecting the optimal path in mixed pedestrian and vehicle environments. This advancement provides crucial support for autonomous delivery vehicles to operate seamlessly on the road.

Furthermore, at the 2024 World Robot Conference, JD Logistics' sixth-generation autonomous vehicle also endeavors to achieve more precise behavior prediction and simulation technology by incorporating large model technology to optimize decision-making strategies, thereby better coping with complex and dynamic road conditions. For instance, JD Logistics' sixth-generation autonomous vehicle can utilize end-to-end models to create a four-dimensional spatio-temporal joint world representation, which directly outputs decision-making and planning results, significantly reducing information loss caused by manually defined module interfaces, effectively enhancing the efficiency and safety of autonomous vehicle operations, and fundamentally reversing the predicament of "rigid thinking."

Meanwhile, relatively complex urban road conditions can severely interfere with the control and execution capabilities of autonomous delivery vehicles. Especially in areas with strong signal occlusion or interference, such as CBDs with high-rise buildings or tunnels, the positioning accuracy of autonomous delivery vehicles will decline. Ensuring accurate position determination and driving control of autonomous delivery vehicles is a pressing issue.

In this regard, in terms of "ears," Neolix actively explores the application upgrade of 5G networks, augmenting the uplink bandwidth capacity of 5G networks, indoor 5G enhancements, and the implementation of 5G-A, leveraging the large bandwidth and low latency characteristics of 5G to clear communication barriers for the large-scale deployment and application of autonomous vehicles.

It is evident that communication solutions are garnering increased attention in the market. Even for brands not specializing in autonomous delivery vehicles, there are professional manufacturers dedicated to resolving current autonomous vehicle communication issues. For example, Quectel and Hypera also provide 4G and 5G series automotive-grade modules for autonomous delivery vehicles, offering robust support for their communication and positioning, ensuring that the vehicles can accurately receive instructions, transmit data, and ascertain their locations.

Autonomous Vehicles Still Need to Overcome the 'High Cost' Hurdle

Overall, driven by manufacturers such as Cainiao, autonomous vehicles have gradually ventured beyond test roads and entered urban public roads. Nevertheless, autonomous delivery vehicles still confront challenges such as urban road rights, significant price disparities, and varying charging models, making it difficult for some small and medium-sized logistics companies to deploy them on a large scale.

Here, let's delve into the future commercial development of autonomous delivery vehicles using Jiushi Intelligence's Z2 model, which may be the most affordable in the industry. Currently, the single-vehicle price of the Z2 model is 39,800 yuan, and it still requires a subscription fee of 6,000 yuan per quarter for intelligent driving services, amounting to a total annual cost of 63,800 yuan. Excluding charging, maintenance, and potential failures, the monthly investment stands at 5,316.7 yuan.

However, assessing such costs in real logistics scenarios reveals that they are not cost-effective.

According to the latest data released by BOSS Zhipin, the average monthly salary of couriers nationwide is 6,626 yuan. Among them, couriers with less than one year of work experience earn an average monthly salary of 5,687 yuan. This implies that the cost of an autonomous delivery vehicle is roughly equivalent to the salary of a courier.

Therefore, autonomous delivery vehicles have not yet reached a point where they can genuinely replace human labor. Perhaps, as time progresses and the initial vehicle purchase cost is amortized over years, the cost investment of autonomous delivery vehicles will be lower than that of human labor. Of course, from another perspective, the emergence of autonomous delivery vehicles is not intended to replace couriers but rather to further enhance logistics efficiency through human-machine collaboration.

From this standpoint, the future of autonomous delivery vehicles remains highly promising. Currently, dozens of cities across the country permit large-capacity delivery autonomous vehicles to operate on roads, and the speed of road rights opening is faster than most people anticipated. With manufacturers such as Cainiao and Jiushi Intelligence continuously iterating technology and introducing L4 autonomous vehicles, it may not be long before autonomous delivery vehicles become a common sight on our streets.

*All images in this article are sourced from the internet