← All writing

June 25, 2026 · Robotics · Mechanical Design

The sumo robot's real job is staying in the ring

A mini-sumo robot has two jobs that sound symmetric and aren't — pushing the opponent out, and not getting pushed out yourself. The second one is the whole design.

A mini-sumo robot has two jobs that sound like mirror images and aren’t: shove the other robot out of the ring, and don’t get shoved out yourself. Spend an afternoon with the rules and it’s obvious which one drives the design. Finding the opponent is easy — a pair of forward infrared sensors handles it. Staying planted while you push something exactly as motivated as you are is the hard part, and it’s a mechanical problem, not a clever one.

So the robot is built to be hard to move. It’s a low, panel-built box — a ground plate, a roof, and four walls, modelled in Fusion 360 and drawn to dimension in AutoCAD — about 200 × 200 mm (the mini-sumo size class) with ~60 mm sides, assembled onto an aluminium base. Low and boxy isn’t an aesthetic. A low center of gravity is what keeps your leading edge from tipping up when an opponent wedges under it, and a closed, rigid box is what stops the chassis from flexing and giving up traction at the exact moment you’re leaning into a push. Every mechanical choice is answering the same question: how do I make myself expensive to move?

The sensing is deliberately the simple half. Forward infrared proximity sensors spot a robot ahead and the drive turns toward it and pushes — sense, steer, shove, a tight closed loop instead of a pre-programmed routine. That’s enough. The sensors don’t have to be brilliant, because the contest isn’t won by the smarter searcher; it’s won by the body that’s still in the ring when the shoving stops. I spent my effort accordingly: a sturdy panel chassis I could cut and bolt to a consistent spec, and just enough sensing to point it at the target.

That’s the lesson I took off this one, and it travels well past plywood-and-IR robots: when a contest is defined by a line you must not cross, the winning move is defensive. Figure out what loses the game — here, leaving the ring — and pour the design into not doing that, before you spend a single gram on doing the flashy thing better. The robot that wins sumo isn’t the one that pushes hardest. It’s the one that’s hardest to push.