Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- By 17 July 2026, time ten cross-base transfers end to end and record dock dwell at each end in minutes, separately from driving time, so the baseline is measured in dock minutes rather than kilometres.
- Before the first scheduled loop, put the receiving window for every arrival hour in writing at the far base, including any arrival between 22:00 and 06:00, and name the person who signs the temperature record on each shift.
- Cap dock dwell at 25 minutes per end for the first four weeks and log every minute over the cap; if the average overrun passes 10 minutes per trip, add a dock hand before adding a second vehicle.
- By 14 August 2026, convert your own carton cube into trips using the 154 kg per cubic metre density, and publish the business case as kilograms moved per 10-hour window, because a cost per kilometre will not survive the first schedule change.
Universal Tech's Chungya Holdings completed a driverless cross-base pharmaceutical delivery test in Chongqing, with a medical-grade autonomous vehicle autonomously moving 4 km between two logistics bases, carrying up to 1,000 kg in a 6.5 m3 box with 200 km range. The zero-emission vehicles will handle routine drug transfers and may extend to hospitals and pharmacies.
The Analysis
Four kilometres. That is the entire test route. Two logistics bases in Chongqing, one medical-grade box between them, and nobody in the driver's seat.
Most people read that as a story about a truck that drives itself. I read it and go looking for the number that is not in the release: how many minutes the box sits at the dock at each end. Four kilometres is ten minutes of driving. A cross-base pharmaceutical transfer is not ten minutes of work. It is a loading window, a temperature record, a signature and a receiving slot, and the driving is the smallest line on that list.
If you move medicines, devices, or anything with an expiry date and a temperature log, this trial deserves an hour of your week. Not because the vehicle is impressive. Because it moves the bottleneck off the road and onto the dock, and the dock is where your promise to a hospital actually breaks.
The facts, kept whole. Universal Tech's Chungya Holdings ran the test through Heping Logistics in Chongqing: a medical-grade autonomous vehicle, 4 km between two logistics bases, 1,000 kg rated payload, a 6.5 cubic metre box, 200 km of range, zero tailpipe emissions. Routine drug transfers first, then hospitals and community pharmacies. Six numbers, and they do not describe the same thing. Exactly one of them is about driving.
Start with range, because it is the number that will be quoted at you. Two hundred kilometres against a 4 km leg is fifty legs per charge on the nameplate, before anyone argues about hills or air conditioning. Energy is not the constraint on this route, and it will not be the constraint on a hospital route either, since a loop inside a city is an order of magnitude shorter than the range. Whatever this vehicle is being bought for, it is not endurance.
Then the box. One thousand kilograms inside 6.5 cubic metres is about 154 kg per cubic metre. That figure is the only bridge between the spec sheet and your own freight: measure your carton cube, divide, and you get trips instead of tonnes. Loading by piece count is how these vehicles get wasted. The count looks fine, one of the two limits is already hit, and nobody at the dock is doing the division for you.
The part that decides whether this works is not on the spec sheet. Custody of a temperature-controlled pharmaceutical shipment moves with a signature and a temperature record, and it moves at a dock, not in traffic. A driverless box still needs a person at each end to break the seal, read the logger, sign, and wheel the cage inside. On a 4 km leg that handover is most of the cycle. The vehicle removes ten minutes in the middle and hands you back fifty at the ends.
So the impact is lopsided, and the dividing line is whether you own a pair of bases. A distributor with two of its own facilities on a 4 km axis can run this as a shuttle and take the driver out of the middle without touching the custody chain at either end. A distributor with one facility gets nothing from a point-to-point link, because the value sits in the pair. For importers this lands after customs release: it changes the domestic repositioning between a port-adjacent DC and a regional DC, and it does not touch the clearance clock by a single minute. For exporters moving pharma out of China, the effect is close to zero until the receiving end has the same setup.
You will feel this in the shift roster before you feel it in the cost line. A scheduled loop has an arrival hour, and the receiving base has to have somebody at the dock at that hour, including 02:00 and 05:00. If nobody is there, the vehicle waits, and waiting is exactly the cost the vehicle was supposed to remove. The first collision usually comes with the pharmacy van departure, not with the accounts.
Here is the thing the announcement does not measure. Everyone will track kilometres and payload. The number that decides the business case is minutes of dock dwell per handover, multiplied by the number of human handovers per trip, and that second number is the one that grows when you extend to hospitals and pharmacies. Each new stop adds a receiving window you do not control, a person who may or may not be at the desk, and a signature that has to land before the box is released. Range scales freely. Receiving windows do not. A pharmacy that accepts deliveries between 09:00 and 11:00 is a constraint the vehicle cannot drive around, and a route with six of those windows has six chances to lose an hour.
Run the arithmetic with my assumptions, not the article's. The article gives 4 km, 1,000 kg, 6.5 cubic metres and 200 km, and says nothing about dwell. Assume 12 minutes of driving each way at 20 km/h inside a base area, 25 minutes of dock dwell at each end, and 10 minutes for the seal, the logger and the signature. One round trip is 84 minutes. In a 10-hour operating window that is seven trips, 7,000 kg. A driver-based run with a shift handover and a mandated break might manage five in the same window, which is my assumption again. Now improve one thing. Cut dwell by 10 minutes at each end and the round trip falls to 64 minutes, which is nine trips, 9,000 kg. Raise average speed from 20 to 30 km/h instead and you save 8 minutes a round trip, which is still seven trips. Ten minutes at the dock beats ten kilometres an hour on a 4 km leg, every time.
That is why this pilot should be read as a dock project with a vehicle attached. Before the loop runs on a schedule, the receiving base needs a named person per shift, a written window, and a place to put the cage that is not the corridor. The temperature record has to be readable by the person signing, not by an app the dock has no terminal for. None of this is exotic. All of it is the difference between a loop that runs and a box that parks.
The counterfactual is ugly and it is common. If the dock is not staffed for the arrival hour, the vehicle waits 40 minutes at each end, the round trip stretches past two hours, trips per shift fall from seven to four, and the programme gets written up as a technology that did not pay back. The technology was fine. The receiving schedule was the failure, and nobody puts a receiving schedule in a pilot report. Good kit dies this way, at 05:40, with nobody at the door. Get the dock staffed, then let the box drive. The parcel still has to be signed for by somebody, and that somebody is the whole schedule.
↑ Back to the key points