Article
Computer vision in robotics
Computer vision on a service robot is used mostly for obstacle detection and understanding the immediate surroundings, not for recognising individuals. Cameras and depth sensors feed the robot's sense of what is in front of it. Any venue installing cameras on a moving machine should establish what is processed, what is stored and for how long.
Written by Hybot technical lead, Technical lead, Hyrcan-Tech · · 5 min read
Where Hybot stands
Hybot's platform does not process camera imagery. Sensing belongs to the robot hardware and its manufacturer, and we integrate through a vendor-agnostic abstraction layer that deals in state, position, battery and tasks. If you need to know what a particular robot's cameras do, that question goes to the manufacturer of that model, and we will help you ask it properly rather than answering on their behalf.
What vision is for on a service robot
Mostly one thing: deciding whether the path ahead is clear, and how far away the thing in it is.
That is a narrower job than the word "vision" suggests. Combined with depth sensing, it produces the behaviour venues actually observe — the robot slowing and stopping when someone steps in front of it.
Scene understanding beyond obstacles — recognising that this is a doorway, that is a queue — helps smoothness. Identifying who someone is, is a different capability entirely and should never be assumed to be included, or excluded, without asking.
The failure cases worth knowing
Three, and they are all physical rather than algorithmic:
- Glass. Transparent surfaces are genuinely hard to detect. A venue with glass partitions is a venue where this needs testing, not assuming.
- Reflective floors. Polished surfaces confuse depth sensing in ways that vary with the lighting at different times of day.
- Low light. Cameras need light. An evening venue is a different sensing environment from the same room at the survey visit.
This is a large part of why a site survey predicts robot behaviour better than a specification sheet does. The sheet describes the sensor; the survey describes the room.
The privacy conversation, which should happen first
A camera on a moving machine in a space full of customers deserves a clear answer to four questions, before installation:
- What is processed on the robot itself?
- What, if anything, leaves the robot?
- Is anything stored, and for how long?
- Who can access it?
Get these in writing from whoever supplies the hardware. A venue that cannot answer them cannot answer a customer who asks, and "the supplier said it's fine" is not a position anyone wants to defend.
Where to go next
Cluster hub: AI in robotics.
Frequently asked questions
What do robot cameras actually do?
- Principally obstacle detection and scene understanding — establishing what is in the path and how far away it is, often combined with depth sensors. The task is to decide whether the way ahead is clear, which is a very different problem from identifying who is standing there.
Do service robots recognise faces?
- Face recognition is a separate capability from obstacle detection and should never be assumed to be present or absent. Ask the hardware supplier directly what is processed, whether anything is stored and for how long, and get the answer in writing rather than in a demo.
What is computer vision still bad at?
- Transparent and reflective surfaces, low light, and scenes very different from its training data. Glass doors and polished floors are genuinely hard, which is one reason a venue survey matters more than any specification sheet in predicting how a robot will behave.
Where this fits
This page is part of AI in robotics: what is real and what is marketing. If you are working through the topic in order, these are the neighbouring pages.
Machine learning in robot navigation
Where learned models genuinely improve how a robot moves through a crowded venue, where classical planning still wins, and what Hybot actually uses.
Edge AI in robotics
Why inference on the robot matters when a building's connection is unreliable, what belongs in the cloud instead, and how Hybot's architecture splits the two.
Are service robots safe around people?
What safety means for a robot sharing a floor with guests, children and luggage — stopping behaviour, speed, handovers, and the venue's own duties.
Take it further
If a question here applies to a venue you actually run, the specifics matter more than the general case.