China's Driverless Delivery Vans Expand as Costs, Permits and Street Conditions Shape Growth

At Shenzhen's Futian Station, Delivery Continues After the Crowds Leave

At Shenzhen's Futian railway station, boxy vans begin another shift as evening crowds thin. They leave a storage area, follow mapped routes and carry goods to a drop point near the platforms. Onboard speakers ask pedestrians to keep a distance of two meters. When someone steps into their path, the vehicles stop.

As of October 22, 2025, this routine captures how autonomous delivery in China has moved beyond demonstrations into daily work. The vehicles operate on designated roads between depots and collection points, fitting into delivery networks with established routes and handoffs. Their job is specific: move goods through a manageable part of a much larger journey.

Couriers load parcels into lockers or cabinets at a depot. The vans transport them to a station curb or neighborhood hub, where customers collect them from a staffed counter or an automated cabinet. By reducing repeated short trips and the time drivers spend waiting, the arrangement can help delivery networks handle dense urban demand.

The scale of that opportunity is drawing manufacturers, logistics companies and city governments into the same market. Cainiao, Alibaba's logistics arm, has moved from research into mass production of driverless delivery vehicles. Company leaders say more than 200,000 unmanned vehicles could enter logistics fleets over the next three to five years if current progress continues.

Expansion still depends on conditions that vary from one district to another. A van needs permission to use the road, a route it can navigate reliably, space to stop and a depot ready to load and charge it. China's progress lies in bringing those pieces together in selected corridors, rather than giving robots unrestricted access to every street.

Why China's Delivery Networks Give Automation Room to Grow

China's position rests on three connected strengths: local policy, enormous parcel demand and a domestic manufacturing base. City governments have created pilot zones for intelligent connected vehicles, introduced safety and data reporting rules, and issued permits for driverless testing within defined operating conditions.

Those conditions are known as an operational design domain. In everyday terms, they describe where and under what circumstances a vehicle is designed to drive itself. A clear boundary gives companies a more predictable setting in which to invest, test their systems and improve them.

Delivery demand supplies the commercial incentive. China's e commerce companies and express delivery networks move billions of parcels each year. When volumes are that large, even a small saving on each package can matter across a network.

Domestic suppliers provide another advantage. Manufacturers of sensors, electric drivetrains and vehicle control software can build and refine systems quickly, at costs that support commercial trials. The combination helps companies turn changes in software or vehicle design into further operating experience.

Programs now extend across many provinces, and dozens of cities have approved autonomous delivery vans with large cargo capacities for public roads. Growth is taking the form of connected operations in selected districts and corridors, with local approvals determining where those networks can reach.

What Level 4 Driving Means on a Delivery Route

Today's robovans target Level 4 automation: the ability to drive without human input within a defined area and set of conditions. When a situation falls outside that operating envelope, a remote operator can assist or take control. The practical goal is reliable service on repeatable routes.

WeRide's Robovan W5 illustrates the technology being developed for that work. Its sensors combine lidar, radar and cameras to observe the surroundings in 360 degrees. Together, they identify vehicles, traffic lights, pedestrians, cyclists and road edges, giving the driving software information for its next move.

WeRide says the W5 can operate around the clock, has the largest cargo volume in its class at 5.5 cubic meters, carries up to 1,000 kilograms and travels up to 220 kilometers per charge. The company says its system draws on nearly 1,900 days of field operations and almost 40 million kilometers of autonomous driving on public roads.

The software plans routes, overtakes when safe, avoids obstacles and docks at loading bays. It can also adjust a route to steer around congestion. These tasks connect driving with the practical demands of a delivery shift: reaching the right loading point and keeping goods moving when traffic changes.

Backup systems are central to the safety design. The W5 uses a drive by wire chassis, meaning electronic commands control vehicle functions, with backup mechanisms for carrying out those commands. Overlapping sensors and emergency braking provide further protection. The aim is to preserve control if a single component fails.

Continuous checks monitor the vehicle's driving state, send information to the cloud and flag unusual conditions for human review. That monitoring links each van to the wider operation, where staff can watch for problems and respond when the vehicle needs help.

The Fleet Platform Is Part of the Product

Fleet operators need more than a vehicle that can steer and brake. Their software schedules delivery tasks, monitors vans in real time and sends software updates over the air, allowing systems to be updated remotely.

Vehicles can also be grouped into convoys, allowing one supervisor to manage several vans at once. This approach helps an operator expand its fleet without increasing staffing at the same rate. It makes the relationship between vehicles and human oversight a central part of the business model.

Cloud analytics examine delivery patterns so operators can position vans where demand is expected. If parcel volumes rise in a district, vehicles can be moved there in advance. Better positioning can reduce turnaround time at depots and help avoid having capacity sit idle in the wrong place.

The operating platform therefore connects several decisions that would otherwise be handled separately: which van takes a load, when it leaves, where it waits and how staff respond to an interruption. The usefulness of autonomous driving depends partly on how smoothly those decisions fit into the delivery network.

How Pandemic Trials Led to Wider Deployment

Contactless delivery during the coronavirus outbreak gave the sector an early push. Beijing based Neolix reported a surge in orders in early 2020 as retailers and food delivery platforms sought to reduce contact between people. Its customers included Alibaba, Meituan and JD.com.

Those trials gave operators experience with remote supervision, maintenance routines and loading procedures. After the immediate health crisis passed, companies retained the elements that worked in regular logistics operations.

Companies with robotaxi experience also brought their software and operating teams into goods transport. WeRide expanded from robotaxis into autonomous buses and logistics, applying related perception and route planning technology to different kinds of service.

Delivery offers conditions that suit current systems: repeated routes, modest speeds and the option to work late at night when roads are less crowded. Those conditions reduce exposure to some of the most complicated street situations while allowing vehicles to complete useful commercial tasks.

The development has unfolded through several stages:

• Early 2020: Neolix reported rising orders for contactless delivery vehicles during the coronavirus outbreak, with Alibaba, Meituan and JD.com among its customers.
• April 2025: WeRide's Robovan W5 received approval to test driverless logistics vehicles in Guangzhou's Nansha District.
• July 2025: The W5 secured the first such permit in Guangzhou's Huangpu District.
• October 22, 2025: Driverless delivery was operating at sites including Shenzhen's Futian station, while companies pursued wider deployment through manufacturing partnerships, logistics contracts and local permits.

Small Savings Per Parcel, Demanding Economics Per Van

Cainiao executives have said autonomous vehicles could reduce delivery costs by one to five Chinese cents per package in the right operating setup. The proposed savings come from longer operating hours, fewer staff, less idle time, tighter route planning and faster depot turnarounds. Electricity also replaces diesel in the vehicle's running costs.

Cainiao's facility in Hangzhou's Yuhang district shows how much work a fleet can handle when routes and parcel volumes fit. It already uses more than 20 unmanned vehicles, each with a loading capacity of around five cubic meters. According to the company, each vehicle can move at least 1,500 packages per day.

Those vans handle more than 30 percent of the depot's parcels, meaning autonomous vehicles already carry more than three in every ten packages there. The example shows how a fleet can take a substantial share of a depot's work when demand is concentrated and the handoffs are organized.

Buying and operating the vehicles still brings substantial costs. Sensors of a high specification, computing units and electric chassis all add to the purchase price. Maintenance and repairs can keep vans out of service, while depots may need changes to make loading, charging and parcel handoffs quicker.

Profitability depends on parcel density, how much of each day a van spends working and how many years it remains useful. A vehicle that completes many short trips, charges during quieter periods and avoids major overhauls has a stronger chance of earning back its cost.

Companies also expect software improvements to increase the amount of work each van completes without raising hardware costs. The strongest setting is a busy route with consistent demand and predictable curb access. In scattered suburban areas with many stops that are not repeated, human couriers retain flexibility that robots struggle to match.

Local Permits Decide Where Vans Can Operate

Guangzhou's approvals show how expansion happens district by district. The W5's April approval in Nansha and July 2025 permit in Huangpu followed local policies intended to accelerate intelligent connected vehicles and new energy vehicles across transport, sanitation and smart logistics.

Each approval opens another defined area for testing or operation. Companies still need to fit their service to local rules and road conditions. Wider commercial coverage grows through these individual decisions, alongside improvements in the vehicles themselves.

On the street, the safety approach is visible. Vans travel at low to moderate speeds, use audio messages to announce their presence and stop for pedestrians. They operate on roads mapped in detail, while remote operators watch for unusual events and can assist when a route is blocked or a traffic signal is unclear.

Insurance, incident reporting and data logging are also part of the operating arrangements. Geofencing keeps vehicles within approved boundaries, while backup systems and human oversight support them as they accumulate more driving experience.

China's approach contrasts with some Western markets, where permits can be fragmented and public road testing faces tighter constraints. Vehicles carrying goods often have a smoother regulatory route than passenger services. Moving parcels between controlled depots and fixed collection points presents a more manageable approval task than transporting passengers from door to door.

Manufacturing Partnerships Meet a Growing Field of Competitors

WeRide is working with Jiangling Motors on manufacturing and with ZTO Express on deployment. The plan is to produce vans designed specifically for delivery at scale and introduce them into large logistics networks.

These partnerships connect the driving system to two requirements for commercial growth: the ability to manufacture enough vehicles and access to a delivery network with enough work for them. Cainiao's move into mass production follows the same broader shift from experiments toward regular fleet operations.

Outside China, Tesla has revealed a concept Robovan intended to carry people and cargo. The name overlaps with WeRide's model, while Starship Technologies previously used it for a support van carrying sidewalk delivery robots.

The shared name covers different approaches, but the operating problems are similar: recognizing complicated traffic, behaving safely near pedestrians, fitting into depot workflows and costing less than a driver and a van doing the same job.

Other Chinese autonomous driving companies also see logistics as an opportunity. Robotaxi programs have accumulated data, software and operations teams that can be redirected toward delivery routes. Reliable service, compliance with local rules and control of costs will shape competition for contracts with express delivery firms and city partners.

Where Human Work and City Infrastructure Still Matter

Robovans are best suited to repeatable loops, concentrated parcel demand and service after dark when streets are clearer. Routes linking depots with railway stations, neighborhood hubs or university campuses offer a natural fit. Vans can also wait outside residential complexes and notify recipients to collect packages during a set period.

These arrangements reduce idle driving and help keep vehicles carrying useful loads. They also organize delivery around places where many packages can be handed over together, rather than requiring the van to solve every part of an individual journey.

Most vans lack robotic arms, so people still load and unload them at depots and hubs. Delivering to every apartment in a tall building remains better suited to human couriers. Automation changes the transport stage while leaving people responsible for important work at either end.

Heavy rain, construction detours and large crowds can delay vehicles or force route changes. Speeds are capped for safety, and operators build extra time into schedules to absorb interruptions. Reliable service depends on allowing for those conditions when assigning work.

City infrastructure can determine whether the system runs smoothly. Charging must fit into depot schedules. Stations and neighborhoods need usable curb space, with clear signs to help pedestrians interact with vehicles. Digital maps require regular updates, and reliable wireless connections support monitoring and data transfer.

China's next stage of growth therefore depends on more than producing additional vans. The vehicles, permits, charging schedules, loading teams and collection points must work together. Where those pieces align, autonomous delivery can make service faster and more predictable. Cainiao's projection of more than 200,000 vehicles over three to five years remains conditional on continued progress, rather than a fixed deployment schedule.

Key Points

• Driverless vans operate daily at Shenzhen's Futian station, following mapped routes, issuing audio prompts and stopping for pedestrians.
• China's expansion rests on local permits, billions of parcels in annual demand and domestic suppliers of vehicle technology.
• Cainiao leaders project that more than 200,000 unmanned vehicles could enter logistics fleets over three to five years if progress continues.
• WeRide says its W5 offers 5.5 cubic meters of cargo space, carries up to 1,000 kilograms and travels up to 220 kilometers per charge.
• Guangzhou approvals in April and July 2025 extended opportunities for W5 testing in Nansha and Huangpu.
• Cainiao says vans at its Yuhang facility handle more than 30 percent of parcels, with each vehicle able to move at least 1,500 packages daily.
• Potential savings of one to five Chinese cents per parcel depend on utilization, maintenance costs and efficient depot operations.
• Human loading, complex apartment deliveries, weather, road changes and charging requirements continue to shape where robovans can work effectively.