A novel supply chain waddles underground
Shenzhen’s Line 2 — among the busiest in the city — is now shared with 41 autonomous delivery bots operated by VX Logistics, a Vanke-owned arm partly held by Shenzhen Metro .
Shenzhen’s Line 2 — among the busiest in the city — is now shared with 41 autonomous delivery bots.,These penguin‑like robots, dodge platform gaps, navigate elevators, and board carriages during off‑peak hours.
These penguin‑like robots, complete with LED ‘faces’, navigate platform gaps, elevators, crowds and board carriages during off‑peak hours, and deliver directly to 7‑Eleven stores spread across more than 100 station outlets .
Equipped with panoramic LiDAR and AI routing systems, they make real‑time decisions based on order volume, train schedules, store location, and footfall . As Hou Shangjie, VX’s head of automation, put it:
“They are specially designed with unique chassis systems that allow them to cross gaps to enter lifts and carriages…They will continue to iterate based on real‑world performance”
“They are specially designed with unique chassis systems that allow them to cross gaps to enter lifts and carriages…They will continue to iterate based on real‑world performance”
Solving the underground logistics puzzle
For station-based convenience retailers, delivering goods used to require street-level carts lugged through busy entrances — a time-consuming, peak-hour challenge. Li Yanyan, manager at a participating 7‑Eleven, explained: “In the past, delivery workers had to park above ground, unload goods, and manually push them into subway stations. Now, with robots, it’s much easier and more convenient”
The bots presently operate outside rush hours, tapping into unused capacity in trains to reduce surface traffic, lower labour overhead and speed deliveries . Distributing around 41 bots suffices to meet peak restock demands for all participating stores, particularly in a system that ferries around nine million passengers daily.
Robotics meets smart‑city ambition
This pilot links with Shenzhen’s “Embodied Intelligent Robot Action Plan” unveiled in March, which targets widespread deployment of service and industrial robots by 2027 . With over 1,600 robotics firms, the city is positioned as a public‑space automation testbed.
National policy also encourages robots in roles from eldercare to hazardous‑environment inspection — blending human labour with mechanised efficiency . For VX Logistics, this is a stepping‑stone to creating a “flexible delivery artery” that extends across metro networks, linking upriver transport modes with automated subsurface logistics.
Next stops: parcels, pharmacy, patient care?
If the trial succeeds, the model could scale beyond snacks to parcels, cleaning supplies — even medical deliveries. Its significance lies in pioneering automation not just at street level, but mid‑route via public transit.
As Li Yanyan observed, tasks once tedious for humans have now become “a routine hop on and off the train” .
As Li Yanyan observed, tasks once tedious for humans have now become “a routine hop on and off the train” .
Foreign observers have noted that this system signals a broader shift — from robots in factories to assistants in everyday life . Soon, commuters in other cities might also share their ride with these fetching, friendly machines.
What it all means
Urban efficiency: The project eases surface‑traffic congestion, optimises underground spaces and trims delivery costs.,Automation in action: Shenzhen showcases a fully integrated, multimodal logistics network that merges AI, infrastructure and public services. Scalability tested: Future iterations may carry mixed parcels and connect with freight systems — framing metro lines as logistical arteries, not just passenger routes.
As robotic supply chains glide through underground rails, we edge closer to a future where public transport is as much a delivery network as a commuting network. Would you feel comfortable sharing your train with a bot that's restocking your local store?
Latest Comments (6)
interesting how they frame "reduce labour overhead" when you know the cost to deploy and maintain 41 of those robots, plus the whole AI routing system and LiDAR. the ROI on replacing a few delivery guys has gotta be super long-term, especially with custom chassis. wonder what the actual TCO looks like.
this is really cool! we just talked about last-mile delivery solutions at our last Cebu.AI meetup, and this totally takes it to the next level with the "middle-mile" in the subway. imagine these little guys helping out with urban logistics here, especially with the traffic. so clever how they use off-peak hours!
The part about these robots handling the "last mile" to convenience stores, that's really interesting for BPO. We're always looking at how automation can cut down on logistics costs, especially in crowded cities like Manila. If these penguin bots can navigate Shenzhen's subway, imagine what they could do for deliveries here, reducing the need for human couriers in certain roles.
This highlights the practical deployment challenges AI faces, beyond just technical capability. The bespoke chassis for gap crossing is a brilliant bit of engineering, showing how systems need to adapt to existing infrastructure. Reminds me of the discussions we've had at the UK AI Safety Institute around ensuring robust physical integration for public-facing AI.
The mention of 41 bots being sufficient for numerous stores, even with 9 million daily passengers, is quite efficient. We have similar challenges with urban last-mile delivery in Paris, and our work at INRIA on multi-agent reinforcement learning for dynamic routing could certainly benefit from examining their path planning algorithms, en effet.
VX Logistics having Shenzhen Metro as a partial owner is a smart play. This kind of infrastructure access, especially in a high-density urban environment like Shenzhen, is a massive competitive moat. In Korea, we see a lot of investment in last-mile delivery robotics, but securing dedicated, integrated routes like this for middle-mile is a tougher nut to crack. It reduces the operational variables significantly. It makes you wonder how replicable this model is without that direct government or transit authority buy-in. That partnership is probably worth more than any tech spec on the robots.
Leave a Comment