# Robot Vacuum AI Obstacle Avoidance, Explained

> How robot vacuums see socks, cables and pet messes: LiDAR, cameras, structured light and time of flight, and why cables are still hard.

- Canonical URL: https://webofmike.com/ai-life/home/robot-vacuums/robot-vacuum-ai-obstacle-avoidance/
- Author: Mike Moore (https://webofmike.com/about/)
- Published: 2026-11-03
- Last modified: 2026-11-03
- Tags: AI Life
- Cite as: Mike Moore, "Robot Vacuum AI Obstacle Avoidance, Explained", Web of Mike (webofmike.com), 2026-11-03. https://webofmike.com/ai-life/home/robot-vacuums/robot-vacuum-ai-obstacle-avoidance/


[AI Life](/ai-life/) / [Home](/ai-life/home/) / [Robot Vacuums](/ai-life/home/robot-vacuums/)

Every robot vacuum box now says "AI obstacle avoidance." The sensors behind that phrase vary a lot, and so do the results. Here's what's actually doing the seeing. Model picks live on the [robot vacuum hub](/ai-life/home/robot-vacuums/).

## The sensors

- **LiDAR tower.** The spinning puck on top maps your rooms. It scans in a plane well above the floor, so it can't see a flat cable at all ([Techracker](https://techracker.de/saugroboter-kabel-erkennung/)).
- **Front camera with object recognition.** An RGB camera plus a trained model that labels things like shoes, socks and pet waste. Roborock says the Saros 10R recognizes 108 object types ([TechGearLab](https://www.techgearlab.com/reviews/smart-home/robot-vacuum/roborock-saros-10r)).
- **Structured light.** Projects a pattern on the floor to read depth. Dreame pairs it with an RGB camera on the X50 Ultra ([Vacuum Wars](https://vacuumwars.com/roborock-saros-10-r-vs-dreame-x50-ultra-full-test-comparison/)).
- **Time of flight (ToF).** Measures distance by how long light takes to bounce back.

The robots that do best in published tests combine a camera with a 3D sensor ([Techracker](https://techracker.de/saugroboter-kabel-erkennung/)). A bumper alone only finds things by hitting them.

## Why cables are hard

A charging cable is a few millimeters thick, lies flat, and is often the same color as the rug. Nearly every model tested does worse on carpet than on hard floor ([Techracker](https://techracker.de/saugroboter-kabel-erkennung/)). Labs also use different cables and layouts, which is why the same robot can ace one test and snag in another.

## How labs score it

Vacuum Wars runs a 24-object course, and the test average is about 16.6 objects avoided ([Vacuum Wars](https://vacuumwars.com/roborock-saros-10-r-vs-dreame-x50-ultra-full-test-comparison/)). RTINGS uses a fixed room with a loosely coiled USB cable on carpet, an extension cord on hard floor, socks, shoes, a bowl and fake pet waste, and scores the worst of three runs ([Techracker](https://techracker.de/saugroboter-kabel-erkennung/)).

## What to do with this

- Look for camera plus 3D sensing on the spec sheet.
- Check the avoidance sensitivity setting in the app. Higher caution protects cables but leaves bigger gaps uncleaned.
- Read results from more than one lab. See the [socks and cables comparison](/ai-life/home/robot-vacuums/roborock-dreame-ecovacs-obstacle-test/) for where labs disagree.

The same camera-plus-model idea shows up in [security cameras](/ai-life/home/security-cameras/local-ai-cameras-no-cloud/), where the question becomes where the video gets processed.