ZEAT

ZEAT is a ground-based cleaning robot for stadium grandstands: industrial design across the robot, its mechanisms, and the system that dispatches it.

Every event leaves three to four tons of trash, and the building has eight to twelve hours to be clean before the next crowd. Today dozens of people close that gap overnight.

ZEAT works that window: sweeping the aisles, reaching onto seats with a folding arm, crossing steps on deformable wheels.

One machine for the floors, the seats, and the steps of a building that refuses to be flat.

Cleans floors and seatsCrosses grandstand stepsCompacts what it collectsReports what it cleanedRuns as a dispatched fleet

The hard part isn’t the trash. It’s the terrain.

A grandstand after the crowd leaves. My own photo, from field research.

Cleaning machines exist, but nearly all assume the one thing a grandstand never offers: flat, open floor. Stepped tiers, narrow aisles, trash on seats.

The design question wasn't a better vacuum. It was a machine that survives the architecture.

On-site field visits during and after events, interviews with a cleaner of seven years and an operations manager of ten, and a questionnaire on spectator littering.

What the Trash Is

Cups and bottles, snack packaging, crumbs and dust, liquid spills. Light, irregular, and split across two surfaces, the floor and the seat. That is exactly the combination existing equipment can't handle.

What Already Exists

Current machines are single-function tools for flat floors. The research frontier held the missing pieces: a patented stair-cleaning concept, published shape-morphing wheels, and YOLO-family recognition. No product had put them together.

Three jobs emerged: recognize trash on both surfaces, collect it including the oversized pieces, and route itself through an unfamiliar seating layout.

In a grandstand, how the machine moves decides everything else it can do.

Seat-Back Rail

Rides the seat backs. Cleans seats well and needs no construction, but it can't cross a step, barely touches the floor, and only fits fixed seating.

Overhead Track

Hangs from installed track. Wide coverage, but it needs permanent infrastructure in every venue, and hardware above the stands intrudes on the one thing a stadium sells: the view.

Ground-Mobile

Harder engineering, but it collects and compacts floor trash, reaches seats with a folding arm, crosses steps on deformable wheels, and adapts to any layout without touching the building.

Both rivals cleaned well. Neither could cross a step. Going to the ground cost complexity and bought the machine the run of the building.

Every Dimension Argues With the Building

816mm long, 496mm wide, 388mm tall, and none of it arbitrary. Wheelbase from the step geometry, height from seat clearance, width from how much floor one pass should clean.

Four wheels, four motors: the body rotates in place, so a machine nearly a meter long reverses inside an aisle it almost fills.

Swallow First, Sort Later

Event trash is oversized, so ZEAT leads with a high-suction intake wide enough to take cups and boxes whole, then compacts them with an internal roller.

Drive and control hardware sits in the wheel housings and floor pan, leaving the middle of the machine empty for collection. The underbody does the fine work: toothed intake conveyor, cleaning roller, twin edge brushes for the space under the seats.

The Arm Is Why It's Not a Floor Robot

Half the trash never touches the floor. A folding arm lifts what the crowd left and drops it through a dedicated inlet, clearing one step's seat and the next step's seat back in the same pass.

Wheels That Trade Shape for Terrain

On flat ground the wheels stay round and roll. At a step, the rim opens into curved spokes that walk the machine up the riser, the obstacle that kills every other cleaning robot in this building.

Sensing

Cameras handle trash recognition on YOLOv5, ultrasonic radar covers obstacles, and an IMU with wheel encoders keeps the machine located on the venue map.

Built

I modeled every component for fabrication, printed across SLA and FDM, finished the surfaces by hand, sprayed the paint, and assembled the body with its lighting. It is an appearance model in the production CMF, and it drives.

A render can hide a weak surface transition behind a camera angle. A painted part on a table in front of a critic cannot.

A single robot doesn't clean a stadium in eight hours. ZEAT is designed as a fleet with a loop.

The Operating Loop

Scan and initialize, take assigned zones, clean the floors, and work the seat rows. Then return to base, empty into a self-cleaning station, recharge, and upload the report.

Dust bag and battery sit under one lid, placed for fast turnaround at the station rather than mid-shift maintenance.

Dispatch

The management app runs the fleet against the venue's calendar: a floor plan shows each zone's load as a heat map, device cards track battery and progress, and scheduling follows the events, so the system knows when the window opens.

The Report

Each finished zone returns the area it covered and photographs of the rows behind it, so the cleaning can be checked in the app rather than on foot.

The Cost I Wrote Down Instead of Hiding

The ground-mobile concept's biggest liability wasn't engineering. A working fleet displaces much of the cleaning crew that does this job today. I kept that line in the evaluation rather than deleting it. A design that automates labor should say so.

What This Project Taught Me

Industrial design here meant negotiating with a building. Every dimension and mechanism answers something the grandstand refuses to change. Screens ask what the user wants. Terrain doesn't negotiate.

What I'd Do Next

The model proves the form, not the mechanisms. It drives, but its wheels don't morph and its arm doesn't lift. Next is functional: build the morphing wheel against real step geometry, validate the arm's reach across seat types, and put the dispatch flow in front of the people whose 3 a.m. problem this is.