Cleaning

The hardest room

Restroom cleaning is the hardest room in facilities work: wet, tight, and the one room a camera cannot enter. It also has a legislated wage escalator behind it. Singapore's Progressive Wage Model floor for an outdoor, healthcare or restroom cleaner runs S$2,325 a month for July 2025 to June 2026, S$2,495 from July 2026, and S$2,835 by July 2028-29, plus a mandatory bonus of at least two weeks. That lands on a sector reported at about 4 percent net margin on a 58 percent labour cost base [single source]. The escalator, not novelty, is why this market is real. The category read vendor by vendor is in Restroom cleaning robots, September 2026.

The standards we design to

IEC 63327, for autonomous commercial floor-treatment machines, is the certification route, with IEC 60335-2-72, IEC 60335-2-79 for steam and high pressure, IEC 60335-2-69 for commercial wet and dry vacuum, and ISO 13482 second edition as the service-robot dossier. ISO 3691-4 and ISO 26058-1 are the wrong hooks: floor-cleaning robots are out of scope of both. Commercially, NEA's Environmental Services Productivity Solutions Grant co-funds up to 50 percent of a qualifying purchase, and MOH's National Cleaning Standards set the cleaning sequence and the time norms the work is measured against — 45 minutes for a common toilet cleaned once a day, 30 minutes for a toilet with a shower, 10 minutes for a single-room toilet.

Camera-free by rule

There is no camera-based occupancy or people detection in the machine at all. Presence comes from a door interlock, a mmWave sensor and the safety lidar; fixture inference runs on the robot; no image leaves it. That is written into the safety file rather than offered as a setting. It is design intent, and we will not market it as a claim until it has been built and audited by a third party.

Cleaning that leaves evidence

The output of this vertical is not a cleaner floor, it is a record: per fixture, timestamped, and shared across a site through Tidewell Crew rather than sitting in one machine's app. It is aimed at a hospital infection-control audit, not at a phone. The robot runs Tidewell Brain in Solo mode, because an interlocked restroom with the door shut is where a site network is at its worst.

Our first targets

Status

Nothing is built yet. The programme runs to month ten and is gated there on one question: whether the robot's cleanliness score correlates with ATP and fluorescent-marker audits closely enough for a hospital infection-control team to accept it in place of a manual visual audit. If it does not, the programme stops. The first site is a hospital ward-floor common toilet block, through CHART and RoMi-H; a facilities-management operator is second. The machine is Tidewell C1.