Independent rankings
Robot vacuums with obstacle avoidance that see a cable, a sock and a dog bowl for what they are: camera models only, each row naming its sensor.
An obstacle sensor is what decides whether you tidy the floor before every run. Infrared sees a wall; a camera with a depth sensor sees a cable and knows to leave it. Weighted on recognition above everything else, with value deliberately at zero. This ranking answers one question, and price is not it.
20 / 111 models evaluated · Updated
Not sure which one fits? Robot vacuum finder
Shop links are affiliate links: a purchase through them may earn RoboVerdict a commission, which never changes a ranking.
The X50 Ultra robot on a plumbed 249 mm dock: 20,000 Pa, legs for a 4.2 cm sill, a LiDAR that retracts to 89 mm, an 80 °C wash, 220 minutes a charge and a 3-year warranty.
Dreame's steam flagship: 40,000 Pa, a 270 mm roller at 18 N that steams at 180 °C, extends 8 cm and lifts 15 mm, legs for 5 cm sills, a retracting LiDAR and a 100 °C wash.
The X60 Pro on a dock that fits one of three pad types per room and doses three detergents: 42,000 Pa, 5.2 cm sills or 10 cm double steps, a LiDAR that retracts to 89 mm, two cameras, a 100 °C wash.
Dreame's roller-mop flagship: 30,000 Pa, a 270 mm roller at 9 to 11 N that slides to the wall and lifts 14 mm, legs for a 4.2 cm sill or 8 cm double step, two AI cameras and a 100 °C wash.
The L60 with legs: 35,000 Pa, ProLeap legs for 8.8 cm double-layer thresholds, a LiDAR that retracts to 89 mm, 280 object types, mops with 40 °C water, and a 6,400 mAh battery for 220 minutes.
The X60 Pro with a plumbed dock only 249 mm tall: the same 42,000 Pa, 89 mm clearance, climbing legs and 18 cm mop reach, with water piped in and out instead of carried in tanks.
Roborock's flagship: 36,000 Pa, twin spinning mops the dock washes at 100 °C, and a solid-state 3D sensor instead of a LiDAR turret.
Dreame's ProLeap legs on a fixed-LiDAR L robot: 19,500 Pa, a 4.2 cm sill or 6 cm double step, an RGB camera with structured light, pads that lift and detach, a 75 °C wash, 200 minutes a charge.
A legged flagship whose dock changes its own mop pads: 30,000 Pa, a 4.2 cm sill or 8 cm double step, pads that extend 45 mm and detach, a 100 °C wash, three detergents and a 16.6 kg station.
The 2025 flagship that introduced Dreame's legs: 20,000 Pa, a 4.2 cm sill or 6 cm double step, 89 mm with the turret down, pads that extend and detach, an 80 °C wash and 220 minutes per charge.
Obstacle avoidance: Twin-camera AI
Best subscore on this list · Navigation: 96/100
Dreame's American flagship: 35,000 Pa, 4.5 cm sills and 8.8 cm double steps, a LiDAR that drops to 79.5 mm, pads that extend, detach and run 40 °C water, an 80-minute charge and a 100 °C wash.
Obstacle avoidance: Camera + depth sensor
Highest on this list · Runtime: 3 h 52 min
Dreame's top vacuum: 42,000 Pa, a turret that retracts to 89 mm, legs for a 5.2 cm sill or 10 cm double step, a mop pad that reaches 18 cm out and detaches on deep carpet, and a 100 °C wash.
MOVA's pad-mop all-rounder: 24,000 Pa, 4.2 cm sills or 6 cm double steps, an 89.5 mm body, pads that swing, lift and detach, and a dock that washes in hot water and dries mop, bin and bag at 60 °C.
The Aqua10 with a track mop instead of a roller: water heated to 45 °C on the robot, 25,000 Pa, legs for a 4.2 cm sill or 6 cm double step, two AI cameras and a 100 °C wash.
A retracting-turret L-series machine: 19,500 Pa, dual anti-tangle brushes, camera plus 3D structured light, pads that lift, extend and detach, a 75 °C wash and 230 minutes on a charge, 89.5 mm tall.
MOVA's roller mop with legs: 28,000 Pa, a roller at 4,100 Pa that extends to the wall and lifts 14 mm, 4.8 cm sills or 8 cm double steps, and a dock that washes at 80 °C, dries, refills and doses detergent.
Obstacle avoidance: Camera + depth sensor
Highest on this list · Dust bag interval: 125 days
MOVA's top roller mop: 36,000 Pa, a roller pressed at 18 N and scraped as it turns, 9 cm double steps, a 96 mm body, and a dock that washes at 100 °C, dries at 70 °C and refills for 1,300 m².
Obstacle avoidance: Camera + depth sensor
Best subscore on this list · Value: 52/100
A Saros-shaped Qrevo: 25,000 Pa, a fully retractable LiDAR in an 80 mm body, 12 N spinning mops left in the dock for carpet, a chassis that lifts 20 mm and takes 4 cm double steps, and a 100 °C wash.
The robot vacuum with an arm: 22,000 Pa, solid-state 3D sensing in an 80 mm body, and a five-axis gripper that lifts socks, slippers and tissues out of its own path.
Roller mop and legs without the retracting turret: 30,000 Pa with a carpet boost, a roller at 9 to 11 N that lifts 14 mm under a guard, a 4.2 cm sill or 6 cm double step, 120 mm tall, a 100 °C wash.
The makers describe five ways a robot vacuum notices what lies in front of it. A camera with a depth sensor names the object and measures where it lies; the depth sensor brings its own light, so the robot still sees the shape with the lamps off. Two cameras judge distance the way a pair of eyes does, so an object the software has never been taught still gets a measured berth. A single camera recognises what it has been taught, a cable, a sock, a shoe, and needs light to do it. Structured light without a camera projects an infrared pattern onto the floor and reads that something a few centimetres high is there, without knowing what it is. Infrared on its own, or no sensor at all, means the robot finds things by bumping into them; those machines are not on this page. Every row below names the sensor according to the maker, and no maker publishes how often it misses a cable.
The camera says what the object is and the depth sensor, a 3D structured-light or time-of-flight unit, says exactly where it lies. This is the group for a home with cables on the floor, a pet, or a robot that runs at night.
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Eufy press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Ecovacs press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Two lenses give depth the way two eyes do: the software names what it recognises and measures the distance to what it does not. The right choice where the floor collects things the software was never taught.
Dreame press image
Obstacle avoidance: Twin-camera AI
Narwal press image
Obstacle avoidance: Twin-camera AI
Dreame press image
Obstacle avoidance: Twin-camera AI
Narwal press image
Obstacle avoidance: Twin-camera AI
Narwal press image
Obstacle avoidance: Twin-camera AI
Narwal press image
Obstacle avoidance: Twin-camera AI
One lens and a recognition model: the robot names a cable, a sock or a shoe and picks a margin for each, and it needs light to do so, which is why some makers add a small lamp. Enough for a lit home where the usual things end up on the floor.
MOVA press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
Roborock press image
Obstacle avoidance: Single-camera AI
Dreame press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
Roborock press image
Obstacle avoidance: Single-camera AI
Eufy press image
Obstacle avoidance: Single-camera AI
Eufy press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
Narwal press image
Obstacle avoidance: Single-camera AI
Roborock press image
Obstacle avoidance: Single-camera AI
Dreame press image
Obstacle avoidance: Single-camera AI
An infrared pattern is projected onto the floor ahead and a sensor reads how objects bend it, so the robot steers round anything a few centimetres high without knowing what it is: a doorstop and a sock get the same berth. These machines are outside the list above, because the page is about recognition and this sensor does not recognise; they are here because the group is the right choice for a home where the floor is cleared before the robot starts and no lens is wanted in the room. The ranking of robot vacuums without a camera orders them together with the bumper-only machines, on cleaning, the map and the dock.
Dreame press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Narwal press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Ecovacs press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Narwal press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Eufy press image
Obstacle avoidance: Structured light
Eufy press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Narwal press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
Dreame press image
Obstacle avoidance: Structured light
Narwal press image
Obstacle avoidance: Structured light
Roborock press image
Obstacle avoidance: Structured light
MOVA press image
Obstacle avoidance: Structured light
The claim means one thing: the maker says the camera recognises pet waste and steers the robot round it instead of through it. It needs a camera, because structured light sees a lump and cannot tell it from a doorstop, and it is the maker's claim; how often it holds is not something any maker publishes. The machines below are the ones whose maker makes that claim. For a home with a dog or a cat, the ranking for pet hair weighs the same recognition together with brush and suction.
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Twin-camera AI
Dreame press image
Obstacle avoidance: Camera + depth sensor
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Photo: ECOVACS ROBOTICS press kit
Obstacle avoidance: Camera + depth sensor
Dreame press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
Roborock press image
Obstacle avoidance: Camera + depth sensor
MOVA press image
Obstacle avoidance: Single-camera AI
MOVA press image
Obstacle avoidance: Single-camera AI
Dreame press image
Obstacle avoidance: Single-camera AI
Mapping is navigation: the sensor on top of the machine measures the distance to walls and furniture and draws the floor plan the app shows, and from that the robot plans its lanes, its rooms and its way back to the dock. Most machines in the catalogue map by LiDAR, a laser that works in the dark; a few map by camera. A robot can hold a perfect map and still drive over a phone cable, because the map is drawn at wall height and shows nothing lying on the floor. The LiDAR ranking answers the mapping question on its own.
Obstacle recognition is the sensor at the front, looking at the floor just ahead of the robot; it is what every row on this page names, and it is scored apart from the map. The two are chosen separately: a LiDAR tower with a bare bumper, a LiDAR tower with a camera and a depth sensor, or cameras doing both jobs. Our guide to how a robot vacuum sees walks through both systems and which machines carry which.
Which robot vacuums we list carry a camera, what their makers state about the images, the incidents reported so far and how to buy one without a lens.
Robot vacuum obstacle avoidance from the makers' own figures: bumper, structured light, cameras and depth sensors, and which machines we list have which.