Survey: Unmanned Car Soldiers Suffer Defeat on the Shanghai Scene

From the perspective of emergency response in disaster relief, unmanned… The car’s contribution Relatively limited, human–machine collaboration currently represents the primary challenge in real-world operations.

Living in a modular hospital, one can… “Ordering takeout.” Inside the makeshift hospital at the Shanghai International Convention and Exhibition Center, patients can enter their location and bed number to place an order, then wait for unmanned delivery carts to bring supplies right to their beds.

The volume of supplies that need to be delivered inside the modular hospitals is enormous, and unmanned delivery vehicles can provide contactless distribution. However, “Contactless” solutions are not an urgent necessity inside makeshift hospitals. What truly calls for contactless delivery is the vast number of locked-down and controlled‑access communities outside those facilities in Shanghai.

Since Shanghai entered a citywide “static management” regime at the end of March, the city’s logistics system has come to a standstill. Subject to The “stay-at-home” policy has left a large number of food delivery riders and couriers stranded inside residential compounds, leading to severe capacity shortages. According to data from the Shanghai Municipal Bureau of Statistics, in March alone, the city’s total number of express deliveries fell by more than 20% compared with the same month last year, while the overall volume of freight transport declined by 6.4%.

In the first quarter of this year, Shanghai’s total retail sales of consumer goods amounted to RMB 438.2 billion—after maintaining a 3% year-on-year increase in January–February, March saw an 18.9% decline, bringing the quarterly total down 3.8% compared with the same period last year. Meanwhile, driven by surging demand amid difficulties in securing groceries, the combined retail sales of the three platforms—Dingdong Maicai, Hema Fresh, and Meituan Maicai—rose by 38.6% in March.
After April, most Shanghai residents relied on government‑distributed gift packages and community‑organized group purchases for their supplies. Both of these channels faced a common challenge: last‑mile delivery. Resident “Staying at home,” delivery to residents’ doors is typically handled by property management staff and volunteers, who lug supplies while clad in airtight protective suits—exhausting work.
On one hand, there is an urgent demand; on the other, a shortage of manpower. Unmanned delivery appears to be a theoretically viable new capacity‑building option. Tech companies including Meituan, JD.com, Newstone, XingShen Intelligence, and Bai Rhino have each deployed their own fleets of autonomous delivery vehicles to enclosed areas such as university campuses, residential communities, and makeshift hospitals.

After After compiling the latest operational data from various organizations and speaking with numerous industry insiders, local volunteers, and residents, it was found that, From the perspective of emergency response in disaster relief, the contributions of unmanned ground vehicles are relatively limited, and human–robot collaboration remains the primary challenge in real-world operations. From a commercial perspective, issues such as mass‑production and deployment scale, cost control, and collaboration between logistics providers and vehicle manufacturers within the supply chain remain critical considerations that the industry must carefully address.

01 How many autonomous vehicles can deliver to Shanghai?

The autonomous vehicles in the modular hospital at the Shanghai International Convention and Exhibition Center are supplied by Neolix. On the evening of April 10, Neolithic’s autonomous vehicles arrived at the makeshift hospital and began operations on April 11. Prior to this, Neolithic had also deployed its unmanned delivery vehicles to several communities in Shanghai.

During the pandemic, “Contactless delivery” is a critical necessity, but for delivery vehicles to truly operate “without human operators,” they must first navigate a lengthy, “human‑driven” process.

First, proactively coordinate with communities and makeshift hospitals that are experiencing capacity constraints, and tailor the number of vehicles deployed and delivery routes to their specific needs.

Secondly, a travel permit must be obtained from Shanghai, vehicles are dispatched from various locations, loaded in batches, and transported to Shanghai by drivers.

Finally, frontline operations staff in Shanghai will be needed to dispatch vehicles into the community, train volunteers, and maintain remote online support throughout the operation.

“The biggest challenge we face is getting vehicles from out of town into Shanghai,” said a Neolithic employee. Say At the time, all highways were completely closed, and the permit‑application process was extremely stringent. A convoy of trucks from Changzhou had to first be taken to a designated highway service area to wait, after which several tow trucks holding Shanghai‑issued permits would arrive. “Making contact.” During the initial handover phase, there were instances of both sides hesitating to take even a single step forward, merely gazing at each other across the line.

According to publicly available information, among the autonomous delivery vehicles currently in operation, internet giants such as Alibaba and JD.com account for the largest share of deployments, while leading manufacturers like Newstone and XingShen Intelligence also hold a significant presence.

An industry operator estimates that, at present, the major companies offering autonomous‑vehicle delivery—whether established tech giants or dedicated self‑driving firms—can deploy a combined fleet of roughly one hundred autonomous delivery vehicles in Shanghai, taking into account production capacity, costs, and operational requirements. “Compared with companies that dispatch more vehicles, they may already have bases or factories in Shanghai or its surrounding areas.”
Taking the Neolithic unmanned vehicle as an example, Shanghai’s Zhangjiang District had already deployed it. More than 90 vehicles were deployed, and the nearby city of Changzhou is home to one of China’s largest manufacturing facilities. According to the aforementioned staff member, upon receiving coordination requests from hospitals and residential communities, the company promptly dispatched over 60 vehicles from Beijing, Chengdu, Yangquan, and Changzhou, which were then transported to Shanghai via freight trucks.

According to incomplete statistics, there are currently more than 500 autonomous delivery vehicles are now in operation in Shanghai.

For most companies, it is necessary to… Under the epidemic prevention policy of “do not leave Shanghai unless necessary, and do not enter Shanghai unless necessary,” even the first step into Shanghai is fraught with difficulty.

02 Staff shortages and efficiency challenges in last-mile delivery.

In an old residential complex in Shanghai, hundreds of supplies must be delivered to residents’ doors each day—entirely by hand. The delivery vehicle is a rather unwieldy flatbed cart; sometimes the packages aren’t sealed tightly and get jostled loose when the cart rolls over uneven pavement. Making repeated trips back and forth, the volunteers have to navigate around flowerbeds and gravel paths, and they’ve even learned to spot every pothole on the neighborhood’s roads—leaving them utterly exhausted after each round. “If there were driverless delivery vehicles, we’d be so thrilled we’d cry.”

Even if they can secure support from autonomous vehicles, the results may still fall short of those achieved by human delivery.防疫要求的 “Vehicles and pedestrians are allowed in but not out.” The “no-contact” policy means that driverless delivery vehicles still face numerous hurdles when entering residential communities.

First, during the onboarding process, there are established criteria for site selection. According to A community notice obtained indicates that, within the gated compound, passenger cars must be parked in nearby spaces equipped with charging outlets, and several sets of steps require the use of ramps for assistance.

A self-driving car technician explained that, With the traditional Compared with robo‑taxis and autonomous buses operating on open roads, delivery vehicles primarily used within closed campuses have lower tolerance for steps and small obstacles, and thus face relatively greater challenges in navigating complex road conditions.
Zhu Yulong, founder of “Automotive Electronics Design,” summarized on Weibo the key requirements for the current deployment of autonomous delivery vehicles in Shanghai:
(1) The community has a population of 3,000 or more and has a need for the delivery of epidemic‑prevention supplies;
(2) The community is authorized to issue epidemic prevention passes;
(3) The community’s roadways are relatively spacious (with a width of 2 meters or more);
(4) Autonomous vehicle companies are required to dispatch operations and maintenance personnel to provide on-site services, and the residential community must verify their volunteer status and issue the necessary documentation.
(5) The community provides staff to jointly learn how to operate the autonomous vehicles, assist the vehicle instructors with management, and handle daily loading tasks.
In addition, most autonomous delivery vehicles on the market are powered by electricity.

Relevant person in charge at Xinshiqi Introduction Unmanned vehicles have minimum technical requirements for road conditions; for example, passing through narrow sections, negotiating tight turning radii, and navigating sharp bends all necessitate tailored design solutions.

In the delivery operations phase, autonomous vehicles do not… “Independence” means that even within enclosed delivery zones, human labor remains indispensable—for routine tasks such as opening and closing doors and disinfection, loading and unloading at parking points, battery charging and swapping, and troubleshooting issues like engine stalling during transit. Several companies, including Meituan and Neolix, have stated that, due to… Under the “vehicles and personnel enter but do not exit” policy, frontline operations staff are unable to gain access, necessitating the recruitment of volunteers. For most autonomous‑vehicle companies, implementing remote cloud‑based control and enabling human–robot collaboration represents the greatest operational challenge. According to A recruitment notice for autonomous vehicle volunteers within the residential community indicates that a single vehicle must… Two to three volunteers are assigned to each task. Several companies in the industry report that, through documentation, video tutorials, or remote connections, they can train volunteers to operate autonomous vehicles in an average of 30 minutes to one hour.

A frontline operations professional in Shanghai explained that the company typically needs about one day to train staff and conduct site visits within residential communities. When factoring in the coverage area of fixed routes, vehicle speed, and maximum payload capacity, the delivery efficiency of autonomous vehicles falls far short of that of human‑powered delivery.

Taking Fudan University’s Handan Campus, where Meituan has designated bike‑sharing stations, as an example, according to a report by The Paper, Meituan has currently deployed a total of… Six “Magic Bags 20” vehicles, primarily used for delivering meals to teachers and students, each have a maximum payload of 150 kg and can handle the delivery of over 200 meal boxes. As of April 10, Meituan’s autonomous delivery vehicles had cumulatively completed more than 10,000 deliveries. According to information on Fudan University’s official website, the North Courtyard Residential Area of the Handan Campus comprises several dormitory buildings. Building 103, each with seven floors and two units per floor, can accommodate over 8,000 students. In contrast to the campus’s massive housing demand, the autonomous shuttle system proves to be a drop in the bucket.

According to the volunteer schedule, an online document provided by volunteers from Fudan University indicates that, on that day, the North District student dormitories on campus… (April 22) We are recruiting 25 volunteers, including 10 who will be specifically responsible for meal delivery services using three-wheeled vehicles, which involve coordinating with the team and assisting drivers during deliveries. Volunteers must be familiar with the building layout and proficient in operating electric tricycles and “small electric scooters.”

Volunteers at Fudan University believe that electric tricycles are more practical than driverless vehicles.

Another Fudan University volunteer Introduction Around the dormitory, there are four Meituan autonomous delivery vehicles, each following a pre-planned route. “A single route covers roughly ten buildings, with two vehicles running on each line, delivering to only two buildings at a time. In other words,” A single delivery vehicle even has to make four or five round trips between the cafeteria and the dormitories just to complete the entire route. “At the same time, he noted that the vehicle speed wasn’t particularly high either: ‘It wasn’t until four electric three-wheelers arrived later that we began to see a noticeable improvement in capacity.’”
Labor shortages and the myriad operational challenges inherent in human–robot collaboration severely constrain the ability of autonomous delivery vehicles to alleviate actual transportation capacity pressures.

03 The future scenarios for last-mile logistics and the associated costs remain unclear.

The unmanned vehicle delivery service deployed to assist Shanghai this time is almost entirely free of charge. For the participating companies, the greater value of contributing to Shanghai’s efforts with autonomous vehicles lies in showcasing their technological prowess and gaining valuable experience for future related operations.

Due to various restrictions and logistical constraints, Shanghai’s epidemic situation has made it difficult for autonomous vehicles to deliver meaningful results, despite existing demand. Several patients interviewed, from different makeshift hospitals, said they had seen the autonomous delivery vehicles but did not use them—and did not need them—because there were enough volunteers to handle basic supply deliveries.

Even setting aside the sudden constraints imposed by disaster relief, from the perspective of future large-scale, normalized operations, last-mile logistics still faces numerous challenges for the autonomous delivery‑vehicle business.

The primary application scenarios of last-mile logistics include delivery services such as parcel and express shipping, food delivery, and mobile retail, mainly covering the segment from courier hubs to door-to-door delivery. “The last 100 meters.” Previously, in addition to couriers delivering packages to customers’ doors, existing options included picking up at collection stations and using parcel lockers. However, a 2017 survey conducted by the China Institute of Smart Logistics—targeting consumers primarily born in the 1980s and 1990s—revealed that, among the most frustrating aspects of last‑mile delivery, “no door-to-door delivery” accounted for nearly half of respondents’ complaints. Entering the market through unmanned delivery and deploying automation have become hot topics of discussion in the logistics industry.

By business segment, the mainstream autonomous delivery‑vehicle companies on the market fall into two main categories: one comprises internet giants such as Alibaba, JD.com, and Meituan; the other consists of autonomous‑vehicle manufacturers like Newstone, White Rhino, and WeRide. The former boasts a vast customer base and direct application scenarios, while the latter possesses mass‑production capabilities. As a result, the former often partners with the latter as their supplier for large‑scale production of autonomous vehicles.

For the separation of upstream and downstream production from application scenarios, before a product enters large-scale commercial deployment, suppliers must carefully assess their order forecasts—taking into account factors such as technological iteration and pilot‑operation scale—as well as cost and capacity constraints. This is a critical calculation that the industry must get right.

Last year, Walmart terminated its partnership with the shelf‑inspection robot company. After five years of collaboration, Walmart canceled its plan to expand 1,000 automated stores and halted its previously aggressive push into automation. According to U.S. media reports, the reason is straightforward: “Walmart found that using robots was less cost-effective than human labor.” This dealt a severe blow to Bossa Nova, the startup, leading to a 50% reduction in its workforce.

In comparison with the cost of replacing human labor, according to JD Logistics’ prospectus, In 2020, the average labor cost per delivery was RMB 5.3 per package, with each courier able to handle 56 deliveries per day. By contrast, adopting autonomous delivery vehicles, by 2021—when accounting for maintenance, depreciation, and operational expenses—the average delivery cost per vehicle is projected to rise to RMB 8.5 per package, while still managing 50 deliveries daily.

As the volume of deliveries continues to grow year after year, and with labor costs remaining largely unchanged, the cost of unmanned delivery is expected to keep declining. According to JD.com’s projections, by… In 2023, autonomous delivery vehicles will be able to deliver more than twice the volume of current human‑powered deliveries, all while incurring a cost of just 1.5 yuan per order. On the production lines for autonomous delivery vehicles, manufacturers such as Neolix and WeRide have established factory facilities across various regions and have already rolled out over a thousand units. This year, they all aim to reach a mass-production target of 10,000 vehicles. As for application scenarios, the largest-scale autonomous vehicles currently in commercial or pilot operation on the market still remain… 300 to 400 units. For the large-scale operation of autonomous delivery vehicles, hardware equipment represents the largest cost component.

This year In March, Xia Huaxia, general manager of Meituan’s autonomous vehicle delivery division, told Titanium Media that unmanned last-mile delivery is currently in the exploratory investment phase. Judging by the maturity of the upstream and downstream supply chains, the cost of key hardware components—such as LiDAR—keeps the per‑vehicle price quite high. He expects that, over the next two to three years, the industry will be able to reduce the per‑vehicle cost to below RMB 200,000, and within five years achieve large-scale commercial deployment.

Meanwhile, in terms of overall vehicle pricing, a divergence has emerged as automakers adopt different technological pathways.

With the participation of the Neolithic unmanned vehicle in the Shanghai aid effort. Take the X3 as an example: the vehicle is equipped with nine cameras, two millimeter-wave radars, four LiDARs, and 14 ultrasonic sensors, with a price tag of around RMB 400,000 in the domestic market.

The latest release this month from WeRide. “Xiaomo Tuo 2.0” has abandoned the “LiDAR‑based approach” favored by mainstream autonomous‑driving companies, opting instead for Tesla’s “vision‑only strategy.” It replaces LiDAR with high‑definition cameras and is priced at RMB 128,800.

Compared with driverless taxis and driverless trucks, driverless delivery vehicles do not have high speed requirements. However, in different scenarios, achieving this currently requires… Level 4 autonomous driving capabilities impose higher hardware requirements—such as LiDAR, cameras, and sensors—for open-road operation than for closed‑park environments.

Therefore, according to the marketing claims… The prevailing narrative of “open roads plus closed parks,” along with the cost‑performance proposals offered by various manufacturers, as well as their ultimate effectiveness and market acceptance, will ultimately depend on large logistics customers scaling up operations to foot the bill.

At present, the key challenge for autonomous delivery lies in forging closer integration between suppliers and logistics providers—aligning their capabilities with specific application scenarios and mass‑production requirements—while ensuring effective cost control and scalable operations across the entire supply chain.