What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
China’s driverless delivery vans are not literally operating like hostile video-game characters—but viral clips do show them struggling with the messy edge cases of public roads. A Neolix X3 bounced across gravel, another vehicle encountered corn cobs, a ZTO robovan became stuck in wet concrete, and a Shenzhen vehicle appeared to continue over vegetables beside a road. The footage is funny because the vehicles move slowly and stubbornly; the underlying questions about safety, supervision and liability are serious.
China’s driverless delivery vans are not literally stealing cars or operating like hostile video-game characters. But viral clips do show why low-speed autonomous logistics is harder in public spaces than it looks on a company demonstration route: a Neolix X3 bouncing across a rough gravel road, a smaller delivery vehicle struggling with corn cobs, a ZTO vehicle stuck in wet concrete, and another robovan continuing over vegetables laid out beside a road.
As an Amazon Associate I earn from qualifying purchases.
Those videos make the machines look comically unstoppable—hence the Grand Theft Auto comparison. The more accurate conclusion is less cinematic and more important: China is deploying autonomous delivery vehicles at meaningful scale, while the systems are still encountering messy edge cases involving unusual surfaces, informal roadside activity, traffic conflicts and imperfect operating environments.
The clips do not establish that the entire fleet is out of control, that the vehicles are routinely unattended, or that any particular robot deliberately ignored an obstacle. They show the uncomfortable gap between a vehicle that can navigate a mapped delivery route and one that can reliably interpret everything humans leave in its path.
#1 Best Overall
- 📦 📬 AUTHENTIC DETAILS: Die-cast metal replicas featuring accurate UPS Mercedes-Benz Sprinter and USPS mail truck designs with detailed logos and markings
- 📦 📬 PREMIUM QUALITY: Constructed with durable die-cast metal and plastic parts for long-lasting play value and collectibility
- 📦 📬 PERFECT SIZE: Each vehicle measures approximately 5 inches in length, ideal for display or hands-on play
- 📦 📬 DUAL PACK VALUE: Set includes two iconic delivery vehicles - the modern UPS Sprinter van and classic USPS Grumman LLV mail truck
- 📦 📬 MOVING FEATURES: Both vehicles feature rolling wheels and opening doors for interactive play experience
What the viral videos actually show
Several different incidents have been bundled together online as evidence of “rampaging” robovans. That framing obscures the difference between getting stuck, continuing through an obstacle, traveling the wrong way and being involved in a collision.
| Incident | What is reported | What it does not establish |
|---|---|---|
| Neolix X3 on a gravel road | A video showed the vehicle bouncing heavily over a rough, potholed path. A report connected the pronounced motion with the vehicle’s physical layout, including two large lithium batteries mounted low in the chassis. | It does not prove a mechanical defect, battery failure or loss of control. |
| Neolix vehicle and corn cobs | A smaller Neolix delivery vehicle was shown encountering corn cobs in its path. | It does not reveal whether the system detected the cobs, classified them as traversable, or had an approved route through that area. |
| ZTO vehicle in wet concrete | A ZTO delivery vehicle reportedly entered wet cement and became stuck. | This is an obstacle and route-management failure, not evidence of a runaway vehicle. |
| Vegetables beside a Shenzhen road | A September video appeared to show a woman trying to stop a robovan from driving over vegetables drying beside a road. | Available reporting does not establish whether the vehicle recognized the vegetables, whether the route was authorized, or whether anyone was injured. |
| Wrong-way travel in Xi’an | The Beijing News reported that a vehicle was filmed traveling against traffic near Kuangshan Road on March 23, 2026. The operator’s regional center said it had forwarded the incident to headquarters and would use vehicle video and backend data to determine the cause. | No final technical explanation was published in the available report. |
| Rear-end collision on Jinghe Bridge | On June 12, 2026, a China Post autonomous delivery vehicle in Shaanxi reportedly braked suddenly, possibly after detecting a slowly moving electric bicycle, and was struck from behind by an SUV. Police said the incident was not a wrong-way-driving case. | Responsibility was still under investigation. The initial police description should not be treated as the final allocation of fault. |
That distinction matters. A vehicle that slowly climbs over an unexpected object looks dramatic in a short video, but it is a different engineering and legal problem from a vehicle that crosses into opposing traffic. A sudden brake can be a conservative safety response, a perception error, a planning decision or a reaction to conditions invisible in the video. Without the vehicle’s logs and an official investigation, the clip alone cannot answer which one occurred.
Why a low-speed robot can look like it is on a rampage
These are generally low-speed delivery platforms, not passenger vans racing through city streets at unrestricted road speeds. Beijing’s 2023 unmanned-delivery rules define the vehicles as platforms equipped with sensors, controllers, actuators, communications and networking systems for automated supermarket, food and parcel delivery. In the Beijing trial framework, they are generally managed as non-motor vehicles and subject to a maximum speed of 15 kilometers per hour. Other jurisdictions can set different categories and operating rules.
Recommended Free Tools
At roughly walking or bicycle speed, an error can become visually absurd. A vehicle may continue pushing against an object for several seconds, slowly climb a low obstacle or remain committed to a route while its wheels lose traction. A conventional driver would often stop, assess the situation and choose a workaround. An autonomous system must detect the object, classify it, predict what will happen if it proceeds, decide whether stopping creates a greater risk, and execute a new maneuver—all within the limits of its software, sensors, maps and remote-support connection.
The central problem is distribution shift
Autonomous systems are trained and tested against an operating design domain: a defined combination of roads, maps, weather, lighting, traffic patterns, communications coverage and expected obstacles. Public streets routinely produce conditions outside that neat envelope.
- Loose or unusual objects: Corn cobs, roadside produce, construction material and debris may not behave like the objects used in testing.
- Unstable surfaces: Gravel, mud and wet concrete change traction and wheel resistance. A route that is geometrically passable may not be physically drivable.
- Informal activity: Vegetables drying beside a road, pedestrians stepping into the route and temporary work areas may not appear in a map or standard road dataset.
- Ambiguous obstacles: A system that stops for every object can block deliveries and create congestion. A system that treats every low object as passable can damage property or become stuck.
- Conflicting traffic cues: A slowly moving bicycle, a vehicle approaching from an unexpected angle or a poorly marked road can trigger conservative braking or an incorrect route decision.
Calling this behavior “the van ignored the obstacle” goes further than the evidence allows. The vehicle may have perceived the object but assigned it the wrong class, detected it too late, judged that it could cross, or followed a route with no safe alternative. The available reports do not release the perception and decision logs for the specific viral vehicles, so claims about a particular sensor failure, software bug or deliberate remote control are speculative.
Driverless does not necessarily mean unsupervised
In ordinary language, “driverless” sounds like nobody is responsible for the vehicle. In commercial autonomy, it usually describes the vehicle’s mode of operation, not the disappearance of human oversight. Beijing’s rules require support for manual near-field control, autonomous driving and remote collaboration, along with remote-assistance platforms and emergency capabilities.
Rank #2
- 📦 AUTHENTIC REPLICA: Highly detailed 1:48 O scale die-cast metal model of the 2019 Mercedes-Benz Sprinter Cargo Van in pristine finish
- 📦 FUNCTIONAL FEATURES: Features working pull-back action motor, opening rear doors, and detailed cargo area interior
- 📦 QUALITY CONSTRUCTION: Built with die-cast metal body, rubber tires, and precise plastic components for enhanced durability
- 📦 REALISTIC DETAILS: Accurately captures the distinctive Sprinter van design with detailed exterior elements and proportions
- 📦 PERFECT SIZE: Ideal 1:48 scale model makes it suitable for display or play while maintaining authentic proportions and details
That can mean a remote operator monitors multiple vehicles, helps resolve an exception or takes a permitted control action when the vehicle reaches a difficult situation. It does not mean that every viral clip was remotely piloted. The reports do not establish who, if anyone, intervened in those specific incidents. The key point is that the safety model may include a human support layer even when there is no person sitting in the van.
This is a real deployment wave, not just a handful of prototypes
The videos matter partly because the vehicles are no longer confined to laboratory demonstrations. The scale figures are not all measured in the same way, and several are company or industry claims rather than independently audited fleet counts, but they point to a substantial expansion.
- Neolix: The company says it delivered its 10,000th autonomous vehicle in September 2025 and describes its products as serving urban and regional delivery. A January 2026 report said the company had deployed more than 10,000 robovans across 300 cities, citing company and industry reporting. That number should be attributed to Neolix and related reporting, not presented as an independently verified count.
- Rino.ai: The company describes itself as an L4 autonomous-delivery business. It says it formed a supermarket-delivery partnership in 2020, launched its R5 Robovan for same-city freight in 2024, and focuses on scenario validation and continuous service. Those are company descriptions and marketing claims, not independent performance findings.
- Zhongshan demonstration routes: A March 2026 report from the Zhongshan government described a local fleet jointly developed by Rino.ai, ZELOS and Neolix, with China Post and SF Express participation. The vehicles began operating on four demonstration routes for express logistics, last-mile delivery and meal distribution. The city said each could complete four to six trips per day and handle 250 to 500 parcels.
- Other operating zones: Xinhua reported in February 2026 that driverless delivery vans were being used at a ZTO sorting center in Inner Mongolia. It also reported that Hai’an in Jiangsu had launched a comprehensive unmanned-delivery demonstration zone covering express logistics, supermarket goods and food delivery.
- Reported national figures: A June 2026 China Daily report, citing China Communications and Transportation Association data, said China’s unmanned-delivery fleet reached 47,000 vehicles in the first quarter of 2026, compared with about 10,000 in 2024, and that more than 100 cities permitted such vehicles on public roads. These are reported trade-association figures, and definitions of “unmanned delivery vehicle,” deployment and permission may differ.
The apparent contradiction is the story: a vehicle can be commercially useful over thousands of routine trips and still behave badly when a route contains wet cement, loose produce or an unexpected traffic conflict. High deployment numbers do not prove that autonomy is solved. They do show why rare edge cases are becoming public incidents rather than private test-track failures.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →What China’s rules require
“Unregulated” is also too broad. China’s national and local frameworks impose operating conditions, although the legal treatment is still uneven from one jurisdiction to another.
The Ministry of Transport’s 2023 trial safety guidance for autonomous vehicles used in passenger and freight transport calls for operation in designated areas after a safety assessment. It requires operators to establish safety systems and risk controls and generally calls for a safety officer to accompany automated freight vehicles. It also addresses vehicle registration, insurance documentation, operational monitoring, incident reporting and suspension of operations when major hazards cannot be controlled.
Beijing’s 2023 rules add specific requirements for unmanned delivery trials, including:
Rank #3
- Officially Licensed by UPS: Features authentic UPS logos and design details, delivering a realistic and engaging delivery truck experience for young logistics lovers.
- Interactive Play Features: Equipped with opening side and rear doors, working lights and sounds, and a free-wheeling forklift with cargo boxes for immersive, hands-on fun.
- Perfectly Sized for Little Hands: Built with kid-friendly dimensions for easy handling and imaginative delivery scenarios—ideal for ages 3 and up.
- Durable Construction with Realistic Details: Made from high-quality materials and designed for rough-and-ready play, with smooth rolling action and detailed finishes.
- Part of the Daron Transportation Toy Line: Since 1990, Daron has been a trusted name in officially licensed transportation toys, inspiring creativity through realistic and educational play.
- operation in designated testing or demonstration areas;
- remote-assistance capability and real-time status monitoring;
- storage of event data;
- insurance and emergency procedures; and
- for commercial demonstration, at least 240 hours or 1,000 kilometers of testing on the relevant roads or area without a responsible traffic accident or traffic violation before a vehicle model qualifies.
Those requirements are safeguards, not a guarantee that every real-world interaction will be handled correctly. They also do not apply identically everywhere. A 2024 Ministry of Transport article said more than 2,000 autonomous delivery vehicles had been deployed in China by the end of 2023 and that more than 30 cities had issued innovation policies, while noting unresolved questions about whether these vehicles should be treated as motor vehicles or non-motor vehicles. Rules on lanes, speed and operating areas were not consistent across jurisdictions.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →That legal patchwork creates practical uncertainty. If a robot is classified differently from a conventional van, it may have different access rights, speed limits and responsibilities. If it brakes suddenly and a following vehicle crashes into it, investigators may need to consider both the following driver’s actions and the autonomous system’s behavior. The Shaanxi incident is a useful example: police reportedly said the SUV failed to avoid the delivery vehicle after its sudden braking, while the reason for the braking and the surrounding conditions remained under investigation.
What the videos prove—and what they do not
They do show
- Autonomous delivery vehicles are operating in varied public and semi-public environments.
- Some systems can struggle with road surfaces and objects that are atypical, temporary or poorly represented in their operating assumptions.
- Low-speed autonomy can still create property damage, traffic disruption and difficult questions about responsibility.
- Deployment is expanding faster than public familiarity with the technology and, in some places, faster than legal standards are being harmonized.
They do not show
- That Chinese robovans routinely operate without remote monitoring or emergency controls.
- That the vehicles are literally “stealing” cars or intentionally attacking pedestrians.
- That every obstacle in a clip was recognized and deliberately ignored.
- That the entire national fleet shares the behavior of one vehicle in one location.
- That viral claims about robots driving through fires, colliding with one another or being remotely controlled are true. Those claims need primary video provenance, vehicle identification, location, date and an official or operator response.
In particular, a social-media video is often missing the information needed to evaluate an autonomous incident: the route authorization, weather, map status, communications state, sensor view, object classification, remote-operator actions, maintenance history and vehicle speed. The most visually compelling explanation is not necessarily the technically correct one.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The bigger question is not whether the vans look silly
The comedy is real. A robot that keeps rolling over vegetables or gets defeated by wet concrete is an irresistible clip. But the safety question is not whether a machine looks awkward; it is whether the operator has designed a system that fails predictably, limits harm and provides a clear recovery path when conditions exceed the vehicle’s capabilities.
A responsible deployment should be able to answer:
- What is the operating design domain? Which roads, surfaces, weather conditions, traffic environments and hours are approved?
- What happens when the vehicle is uncertain? Does it stop, pull over, request remote help or continue at a reduced speed?
- Who can intervene? Is there near-field manual control, remote assistance, an emergency stop and a staffed monitoring operation?
- What data is retained? Can investigators reconstruct the vehicle’s speed, perception, braking decision, route and communications state?
- Who is accountable? Are the operator, vehicle maker, remote supervisor and other road users’ responsibilities defined before a collision occurs?
- How is a blocked vehicle recovered? A delivery robot stuck in wet cement is not merely a customer-service problem if it obstructs traffic or requires people to enter a dangerous roadway.
Those questions are more revealing than the label “AI” or the word “driverless.” A system can be impressive in its normal environment and still need strict geographic limits, human escalation and rapid suspension when an edge case exposes a hazard.
Free tools Windows power users keep installed
One-click scans. No signup required.
Bottom line: a fast-growing experiment with very visible failure modes
China’s autonomous delivery vans are not literally rampaging through cities like Grand Theft Auto. The viral clips show low-speed robots encountering the untidy physical world: potholes, gravel, wet cement, loose food, roadside commerce and unpredictable traffic. Some incidents are embarrassing; wrong-way travel and sudden braking are more serious because they implicate traffic safety and accountability.
The fairest reading is that China has moved autonomous delivery from isolated prototype work into large-scale commercial and demonstration operations before every public-road edge case has been solved. That is neither proof of technological collapse nor evidence that the videos are meaningless. It is a reminder that autonomy is judged not by how smoothly a robot handles its planned route, but by how safely and transparently it responds when reality stops looking like the map.
Rank #4
- Authentic & Official: Realistic replica of the UPS Package Car featuring certified logos and markings, officially licensed by United Parcel Service for true-to-life accuracy.
- Purpose & Play: Fun pullback action lets kids race the truck forward—no batteries needed—perfect for imaginative delivery missions and hands-on play.
- Perfect Size: Measures approximately 5.5” L x 2” W x 2.5” H, ideal for small hands, easy storage, and on-the-go play.
- Quality Materials: Made from durable, lightweight plastic designed to withstand active play while remaining safe for children ages 3+.
- Trusted Brand: From Daron, a leader in realistic vehicle replicas that inspire creativity, learning, and fun through authentic design.
Frequently Asked Questions
Are China’s delivery vans completely unsupervised?
No. “Driverless” describes the vehicle’s operating mode, not necessarily the absence of human oversight. Beijing’s rules require autonomous operation to be supported by remote collaboration, status monitoring and emergency capabilities. The available reports do not establish whether a remote operator intervened in any specific viral clip.
Are the autonomous delivery vans actually rampaging through cities?
The available evidence does not support that broad claim. Individual vehicles have reportedly encountered unusual surfaces, obstacles and traffic situations, but those incidents are different from one another and do not prove that the entire fleet is out of control.
How fast do these delivery robots travel?
Beijing’s 2023 trial rules set a maximum speed of 15 kilometers per hour and generally treat unmanned delivery vehicles as non-motor vehicles for that trial framework. Rules and classifications can vary elsewhere in China.
What happened in the Shaanxi delivery-vehicle collision?
On June 12, 2026, a China Post autonomous delivery vehicle reportedly braked suddenly on the Jinghe Bridge in Shaanxi and was rear-ended by an SUV. Police said it was not a wrong-way-driving incident, but the reason for the braking and final responsibility were still under investigation in the available report.
How many autonomous delivery vehicles are operating in China?
Reported figures vary by definition and source. Neolix says it delivered its 10,000th autonomous vehicle in September 2025. A June 2026 China Daily report citing trade-association data put China’s unmanned-delivery fleet at 47,000 vehicles in the first quarter of 2026. These figures should be treated as attributed company or industry claims rather than one independently audited count.
The Bottom Line
The videos are evidence of brittle edge cases during a rapid deployment wave—not evidence that China’s delivery vans are literally uncontrolled. They should be judged incident by incident, with the vehicle’s logs, operating permissions, remote-support records and official investigations separating a stuck robot from a genuine traffic-safety failure.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
