Restaurants
Service robots for restaurants
A restaurant service robot carries food, drinks and bills from the kitchen to a numbered table and collects used dishes on the way back. Hybot robots take those jobs from a queue that is re-scored every two seconds, so the nearest robot with enough battery and the right priority takes each run without a human dispatching it.
What the robot actually does in a restaurant
Hybot's task model has four job types and all four map onto a dining room
directly: DELIVER_ORDER carries food and drinks out, GET_ORDER brings the
robot to a table to take an order, DELIVER_BILL runs the bill, and
COLLECT_DISHES clears on the way back. A task can also be a multi-step
workflow with a confirmation waypoint — the robot waits at the table until
somebody confirms the step before it moves on.
Tasks arrive from three places: staff create them in the admin interface, the
guest raises one from the table QR page, or a kiosk order generates one
automatically. A guest-originated request is not a second-class queue; it
becomes an ordinary runtime task marked source="customer" and competes for a
robot on the same terms as anything staff created.
How the fleet decides which robot takes a run
Every two seconds the platform re-scores the queue and assigns each waiting job to the best available robot, weighing three things together: the task's priority, the robot's distance to the pickup point, and its remaining battery. Nobody dispatches. A robot below its battery threshold takes itself to charge instead of accepting work, and a robot marked inactive keeps broadcasting its state but is skipped for assignment.
This is the part that decides whether a second or third robot is worth buying. Read how Hybot decides which robot does what for the assignment loop in detail.
The guest-facing layer
A restaurant deployment is not only robots. Each table's point of interest
carries its own QR token, so /c/<token> opens a page that already knows which
table the guest is sitting at — nobody types a table number. From there a guest
can call a robot, call a waiter, open the full order menu or answer a survey,
and see the live task queue with their own request highlighted and a position
indicator reading "task 3 of 7".
The order menu itself is the substantial part: multilingual with automatic right-to-left layout, categories with daily time windows and weekday rules so breakfast disappears at 11:00 on its own, items with option groups, allergens, prep time and combos, live best-seller badges computed from real sales, and bulk import and export through Excel so a translator can be handed a single-language export and hand it back. The full picture is on what your guests actually touch.
Measuring whether it worked
The platform tracks, per robot, lifetime and daily task totals, lifetime and daily distance totals, and time in project. Those are the numbers a restaurant should judge the deployment on, because they are measured rather than modelled: runs completed per service, distance the robot covered instead of a person, and how both changed between week one and week eight.
Surveys add the other half. The default question bank ships in English, Persian and Arabic and has two groups — Restaurant & Food, and Service Robot — so the same form that measures food quality also collects what guests think of the robot. Submitting on the robot's own body display auto-confirms the robot's waiting step, so answering the survey releases the robot without anyone tapping DONE.
The method is worked through in how to calculate the return on a service robot.
Deployments
Hybot runs service robots in a fine-dining venue in Dubai, where the robots carry orders from the pass to the table and take used covers back to the wash. The write-up of that room — service robots in a fine-dining venue in Dubai — describes what the floor needed, how Core RMS was configured and why no performance figures are published for it yet. It is a description of a working deployment, not a results claim; when figures are exported from the fleet record and agreed with the venue, they will appear there rather than here.
What gets checked before a quote
- Aisle width between the tightest pair of tables, and whether trolleys already pass there
- The path from the kitchen pass to the room, including any door that must be held open
- Where a charging dock can sit without blocking a fire route
- The number of floors, and whether a lift is in scope
- Typical covers at peak, which is what sets how many robots earn their keep
All of that comes out of a site survey, which is why every route here ends in a request for a quote rather than a price.
Related reading
- What is a service robot? — the definition, and where restaurants fit in it
- How many robots does a restaurant need?
- Service robots for hotels — the same fleet, a very different building
Frequently asked questions
Do service robots replace waiters?
- No. A Hybot robot takes the carrying: food running, drink running, bill delivery and dish collection. It does not read a room, handle a complaint or upsell. In practice it removes the walking from the job and leaves the parts of waiting that need a person.
How does a robot know which table to go to?
- Every table is a named point of interest on the floor plan and every point has its own QR token. A task created for table 12 resolves to that point's coordinates, so staff and guests refer to the table number and never to a position.
What happens when the dining room is full of people?
- The robot navigates against a stored map of the fixed structure of the room and uses its own sensors for everything that moves. It slows and yields rather than pushing through, which is why aisle width matters more than floor area when a venue is surveyed.
Can guests order without downloading anything?
- Yes. A guest scans the QR code on their table and the menu opens in their phone browser, or they order from the kiosk menu on the robot's own screen. There is no app to install on either route.
What does a restaurant have to change before robots arrive?
- Usually nothing structural. The floor is surveyed, aisle widths and charging locations are agreed, and the floor plan is aligned to the robot's map once. After that the layout can change and only the points of interest need updating.
Where this fits
This page is part of Smart robotic solutions by industry. If you are working through the topic in order, these are the neighbouring pages.
Service robots for hotels
Hotel service robots carry items between the front desk, floors and rooms, using stored multi-floor maps and named points of interest rather than a person walking.
Service robots for healthcare
Healthcare delivery robots move supplies, samples and meals between named points on a ward so clinical staff spend less of a shift walking corridors.
Service robots for retail
Retail service robots carry stock between a back room and the shop floor, guide customers to a named location and run brand video on screen between tasks.
Service robots for offices and workplaces
Office service robots handle internal deliveries, refreshment runs and visitor guidance between named points, and show information on screen between tasks.
Tell us about the venue
Pricing depends on unit count, layout and integration work, so we quote against a real site rather than publishing a package.