
Spent an afternoon watching Huanlesong 2: it can do work, just not work for me
In our factory, from warehouse to line-side, we've always relied on people pushing material carts — two people per shift walking back and forth. I've been wanting to find a cheaper automated delivery solution. Last month a friend who does chain restaurant equipment said their stores got Pudu's BellaBot 2 and asked if I wanted to check it out. I mainly went to see whether this navigation and scheduling could be moved to workshop material delivery — the food-serving part was secondary.
At the store, 2:30 PM, off-peak. The clerk first showed me mapping. This machine uses VSLAM+, navigating by camera, no need to stick magnetic strips or reflective plates on the floor. Mapping is just pushing it around the store once, and lines slowly grow out of a gray background on the screen. I'm familiar with this — the SLAM navigation carts in our workshop work the same way. The difference is restaurant walls are straight, while our workshop walls have material boxes piled against them.
Marking table positions isn't complicated — tap on the interface, assign a number, same idea as recipe management. Three-layer tray, quick-release; the clerk said when washing you just lift the whole thing off and rinse it. Total payload 40 kg, 13 kg per layer, can be expanded up to seven layers. I watched it deliver four tables' dishes in one trip, tray pretty full — this load capacity is enough for a restaurant.
But when I actually pushed it around for two trips, problems showed up.
First is the floor. Near the store entrance there's a terrazzo section, then tile inside, with a threshold of about two centimeters in between. It slows down going over the threshold, and once it even stopped to re-localize. Our workshop floor is way worse — oil, water, iron filings, and pits pressed by forklifts. I can't vouch for this one. The spec sheet says "better environmental adaptability" — sounds nice, but environmental adaptability means different things in a restaurant versus a workshop.
Second is the aisle. The store aisle is just over a meter wide — two people squeezing past is a struggle. It yields to people, but slowly — it stops and waits for them to pass rather than going around. In a restaurant that's fine, guests aren't in a hurry over one minute. On a production line it's a takt time problem — nobody at line-side waiting for material wants to wait three seconds while it thinks.
Third is recharging. It docks itself, and the first time it took two attempts to get in; the clerk said the marker was stuck slightly crooked. This is the same principle as line changeovers — if the baseline isn't right, everything after is trouble. I asked how long it takes to change the table map; he said adding or removing tables takes a few minutes, but changing the layout means re-mapping the whole thing. That's still a small scenario — when we switch vehicle models, all workstations move. I can't accept that kind of time.
On battery life, the spec sheet says 12 hours fully loaded, 15 hours empty, 20Ah lithium iron phosphate battery, about 4 hours to full charge. I couldn't follow it for a whole day, but I watched it run for over three hours in the afternoon and the battery didn't drop fast. I believe that data — running a restaurant all day with one recharge in between is standard.
There were things I thought were done right, too. First, obstacle avoidance stability — guests walking back and forth, kids running, it never emergency-stopped or hit anyone, decelerated very smoothly. Second, the quick-release tray — this detail is really practical. The biggest fear in food service is cleaning hassle; being able to lift the whole thing off and rinse it is what makes staff willing to use it every day. Third, it doesn't wander — it won't go outside its defined area. That sense of boundaries matters in crowded places.
What it really saves is that people don't have to keep watching it — more concrete than saving a bit of walking time.
So I need to be clear about the conclusion. What it suits: restaurants, hot pot places, canteens — places with fixed points, flat floors, predictable foot traffic, especially places where waiters carry plates until their backs ache. This machine can genuinely take over part of the work. What it doesn't suit: production lines like mine — can't keep up with takt time, poor floor conditions, and changing layouts is costly. For in-factory material delivery, you still have to look at the industrial-line models — protection rating and chassis structure are a different design philosophy.
On the way back I was thinking: the machine itself is mature — mature in that it knows where it should work. Forcing it into a workshop is my problem, not its.
Physix Frontier