Cobotics: definition, principles and how it applies to floor cleaning

Cobotics is the name for a robot and a human operator working together on the same task, in the same space and at the same time, each taking on what it does best. The word contracts "collaboration" and "robotics". It is the opposite of robotics as replacement, where the machine takes the person's place: in cobotics the machine takes the repetitive, punishing share of the work, and the person keeps what calls for judgement, adaptation and a presence on site.

The term comes from "cobot", coined in 1996 by J. Edward Colgate and Michael Peshkin, researchers at Northwestern University, for a device designed to handle objects in direct collaboration with an operator, originally on car assembly lines. The idea grew up first in manufacturing, with collaborative robot arms that share a workstation without a safety cage, then in services, with autonomous mobile robots that share people's space: hospitals, warehouses, hotels, shops, and floor cleaning.

This page is our reference on the subject. It defines the three forms of cobotics, separates cobotics from robotics as replacement, sets out how the work is divided between the operative and the robot and then between the robots of a single fleet, applies it sector by sector, deals with jobs and safety, and ends with a glossary and a set of common questions. To see the principle applied to a real site, our how it works page describes the course of a contract; for the machines themselves, see our robots and the guide to the robots on the market. The machines we deploy under contract come from the Pudu range: we are an authorised Pudu distributor.

Robotics as replacement thinks in headcount: a machine takes a person's place, and the business case is built on the hours removed. Cobotics thinks in tasks: you break the work down, you give the machine what it does better than a person (covering a large open floor, at night, as consistently on the hundredth pass as on the first) and you keep for the person what a person does better than a machine (seeing what has changed, deciding, finishing, putting right). The same robot can serve either logic; what separates them is how the work around it is organised.

The difference shows in the result. All-robot gives you sites that are clean in the middle and dirty at the edges, zones skipped because a carton was left in the way, toilets forgotten because no robot goes there, and a standard that rests on the machine alone. Cobotics gives you open floor held by the robot and sensitive points held by the operative, a human check on every shift, and a service that carries on when the machine stops. All-human, finally, gives you a standard that depends on how tired the team is and how many of them turned up, and a level of physical strain the industry can no longer recruit for.

On a site cleaned the cobotic way, the split is explicit and written into the cleaning schedule. The robot has four roles: to cover the floor (scrubbing, sweeping or vacuuming, depending on the machine) in the agreed time slots, on its own; to repeat that pass identically, at the planned frequency; to log every pass (area, duration, zones not done, alerts), which is what gives traceability; and to flag what it cannot do, a blocked zone, a full tank, an anomaly, so that the operative can step in.

The operative has five roles, and not one of them is secondary. Preparation: clearing the aisles, picking up what has been left lying about, moving what blocks the route, checking the docking station. Edges and finishing: skirtings, corners, the foot of racking and furniture, thresholds, stairs, toilets, changing rooms, cluttered areas, everything the robot does not do or does badly. Checking: a human look at the result on every shift, with the quality control checklist, and immediate rework of anything that is not right. The exception: the accidental spillage, the liquid on the floor, the area under building work, the request from the client. First-line care of the robot: emptying and refilling, changing a brush or a squeegee blade, cleaning the sensors, reading the reports, calling maintenance. The split is sized with the fleet-sizing simulator.

The operative's skill remains the skill of the trade: floor cleaning techniques, choosing the right product, reading a floor, following the protocol for a high-risk area. The robot does not replace that; it takes away the part of the job that called for no skill, only stamina.

Cobotics does not stop at the operative and robot pair. On sites of several tens of thousands of square metres, several robots work together, and their cooperation follows the same principles: each to what it does best, in a defined order and on defined zones. An autonomous sweeper runs before the scrubber-dryer, because scrubbing an unswept floor clogs the brushes and the tanks; an autonomous vacuum takes the office carpet while a multifunction robot scrubs the hard floors of the circulation areas; a large scrubber-dryer holds the aisles of a warehouse while a compact model holds the picking zones.

Three building blocks make that cooperation possible. The fleet planner, which allocates zones, times and docking stations between the robots and keeps them out of each other's way. Building integration, lifts, automatic doors and access control, which lets a robot change floor or zone on its own; manufacturers increasingly offer it, but rarely document it. And interoperability between brands, still in its infancy in cleaning, whereas logistics has had the VDA 5050 standard since 2019 to make transport robots from different manufacturers talk to a single control system. Of the 106 models we track, most manufacturers publish nothing about lift integration or about their API; it is one of the data points our guide to the robots on the market records, and the subject of a review we will publish. Our study of cleaning robot interoperability puts a figure on what manufacturers publish for each. The 2026 Observatory consolidates those figures.

For the team, robot × robot cobotics changes the scale, not the principle: the operative supervises a fleet instead of a machine, reads a dashboard instead of a report, and keeps the same roles of preparation, finishing, checking and exception.

Jobs. Cobotics is not there to cut jobs, it is there to make the job sustainable. The cleaning industry can no longer fill its vacancies, in particular at night and on part-time contracts, and physical strain is the first reason: postures, carrying loads, split shifts. Handing the open floor to the robot takes the hardest hours out of the shift, and the operative grows into supervision, quality control and the relationship with the site. Our article on the labour shortage in cleaning covers that context.

Training. A few hours are enough for an operative to run a robot day to day: starting, stopping, moving it, docking it, reports, routine alerts. Setting up the cleaning schedule is done by the distributor. The real learning is organisational: deciding, in the cleaning schedule, what belongs to the robot and what belongs to the operative, and revising that after the first few weeks.

Safety. A cobotic cleaning robot works among people. Recent models detect people and obstacles with LiDAR and cameras, stop and steer around, announce themselves with light and sound, limit their speed and carry an emergency stop. IEC 63327 sets the safety requirements for autonomous floor treatment machines in commercial use; few manufacturers publish their conformity, and it is a data point we record model by model. On site, safety is also a matter of organisation: zones and time slots, signage, instructions to occupants, and the presence of the operative. Our 2026 safety transparency barometer gives the figures, manufacturer by manufacturer.