Three boys in school uniform wheeled a robot into a newspaper office: about the size of a large suitcase, four wheels, and explained spec by spec. Redwan Abrar, Samin Yasir Siddiqui and Minhaz Islam, all in class nine at Adamjee Cantonment Public School, built it. Its name is Deligo and it delivers things without a person.
What it does
Given a destination, Deligo follows a marked lane along a road using a camera and GPS, carrying a box of up to a foot square — food, groceries, a parcel, urgent medicine. On arrival it will not open for just anyone: the recipient confirms identity with a one-time password or a card swipe. When its 20,000 mAh battery runs low it returns to its own charging station and plugs itself in. It reaches 15 km/h on a straight road and slows to 6.5–7 km/h on bends. Total cost: about Tk 1 lakh, raised and spent by three fifteen-year-olds.
Why these three choices are the right three
What makes this more than a school project is which problems the team went after. These are the three that commercial delivery-robot companies spend heavily on, and the students picked all of them.
Navigation, and why the camera matters more than the GPS. Consumer GPS is accurate to a few metres. That is fine for telling a car which road it is on and useless for keeping a knee-high robot on a footpath, because a few metres of error puts it in traffic. So position has to come from what the robot can see. Following a marked lane with a camera is the pragmatic version of that: instead of knowing where it is on a map, the robot knows where it is relative to a line it can observe directly. Commercial systems do a more elaborate version with lidar and pre-built maps, but the principle is identical — the satellites tell you roughly, and the sensors tell you exactly.
Handover, which is a commercial problem disguised as a technical one. A robot that arrives and opens is a box of food on a street. The entire economics of unattended delivery depends on the goods reaching the right person, because the operator carries the loss otherwise. An OTP or card swipe is precisely what commercial fleets use, and choosing it means the team understood that the delivery is not complete when the robot arrives.
Self-charging, which is what separates categories. This is the least glamorous feature and the most important. A machine that needs a human to plug it in is not autonomous; it is a remote-controlled trolley with a person on call, and the labour it was meant to save reappears at the charging point. Returning to a dock and connecting is also genuinely hard: the robot must navigate to a specific spot and then align its contacts within a tolerance far tighter than anything else it does all day, repeatedly, without damaging the connector. Fleet operators treat docking reliability as a core metric for exactly this reason.
What Tk 1 lakh does and does not buy
Roughly eight hundred dollars, against commercial delivery robots that cost tens of thousands. The gap is mostly sensing. A commercial unit carries lidar for three-dimensional obstacle detection in any light, redundant cameras, and compute to run perception models in real time. Deligo is doing a constrained version of the task with a camera, which works on a marked lane in reasonable light and will struggle with an unmarked surface, heavy rain, or a dog.
That is the correct trade for a prototype. The students did not build a worse version of a commercial robot; they built a working version of the easiest useful case, which is how the field itself started.
The constraint no engineering solves
Worth saying plainly, because it applies to every wheeled delivery robot on earth and not only this one.
These machines need a continuous, reasonably flat surface with dropped kerbs and no permanent obstructions. The operators that have succeeded deploy on university campuses — flat, mapped, privately managed, low traffic, with a predictable population and someone to call when a robot is stuck. Those are not incidental conveniences. They are the operating envelope.
A footpath that is interrupted every thirty metres, parked across, used for trading and broken at every kerb would defeat any wheeled robot ever built. That is an infrastructure fact rather than an engineering failure, and no amount of cleverness in the robot changes it.
Why a school project deserves the attention
The team took a silver medal in the Innovative Social Sciences category at the national round of the World Invention Competition and Exhibition and a place in the world finals in Malaysia.
But the result is not the point. University-level robotics in Bangladesh is strong — rover teams placing in the top ten worldwide, underwater vehicle teams — while school-level work is thin, for want of equipment and mentors rather than ability. Three students with a lakh of taka and a school workshop have just demonstrated how small that gap actually is. Whatever happens in Malaysia, the robot works, and the judgement about what to build was better than the budget.
