For some of us growing up, our vision of the future was filled with flying cars and robots that did our work for us. While the hovering sedans of the Jetsons have yet to materialize, robot vacuums—on the other hand—have arrived. And while they don’t look like George’s Rosie, these wafer-shaped, dirt-sucking automatons have come of age. Costing from a bit under $200 to over a thousand, they boast a wide range of features and technology. Given that there’s a lot going on under the hood, so to speak, we’ve tested the most promising ones available to get a fair gauge of their performance and determine the best.
Best Robot Vacuums
What You Need to Know About Robot Vacuums
Keep Your Regular Vacuum
First, let’s get this out of the way: You’re probably not getting rid of your current vacuum. In most homes there will still be some places a robot won’t reach, like in between narrowly spaced furniture legs. And there are some types of dirt/debris that can overwhelm the vacuum filter—these small machines have, not surprisingly, small filters. Fine dust, fireplace ash, flour, corn starch, and powders, even in moderate quantities, can clog a filter and affect performance. If you drop a bag of flour in the kitchen, it might be best to scoop up the majority of it and let your robot finish the job. (FYI, we found the most efficient way to clean the small filters was to use our shop vac with a pointy crevice tool.)
A robot vacuum can absolutely make life easier, cutting down on the time you spend cleaning. Set up on a schedule, it can help the house look and stay cleaner on a day-to-day basis. So when the time comes for you to take over, there’s only some touch-up required. Of course, if there are no kids or pets, that may not even be necessary.
As we tested these vacuums, colleagues and friends invariably asked if we tested them with dog poop. (There are plenty of videos online documenting the ensuing devastation when a robot vacuum trundles over a pile.) Except for one of these vacs, we did not. Until recently, there wasn’t a model that we’ve found capable of identifying pet accidents. But the Roborock below now features such a dedicated pet-mess avoidance system. Other companies have promoted the fact they’re developing this feature, but only Roborock has made it to market so far.
Things to Watch Out For
When using a robot vacuum for the first time, stick around to monitor what it’s doing. There are a few things to keep an eye on. Shoelaces, strings, carpet fringe, and thin or lightweight clothing can all get pulled into the brushes or rollers. The Roborock uses two cameras and artificial intelligence (AI) to identify many common “problem obstacles” the vacuum may encounter and avoid them. Some units have sensors that will stop them when the brushes meet resistance, and some will even reverse, spitting out whatever they pulled in. But, to avoid damage to the vacuum and whatever it might try to suck up, you should keep the area clear. Additionally, be aware of pet food/water bowls, plant stands, or other things that might be “tippy” until it’s clear how a vacuum will interact with them. Models that leverage bump-and-go navigation, like the Eufy G30 Edge, may knock things a little harder, so it’s a good idea to set up boundaries around those water bowls. Lastly, if kids have toys with small parts, like Lego bricks, those can easily be picked up by a robot vacuum. Many models have features to create exclusion zones, so make use of them or employ your own strategy to avoid all these little hazards that could jam up the vacuum.
Navigation is key to how well a robot vacuum performs its job. Suction matters little if the vacuum doesn’t pass over every inch of the floor. Just about all models employ three basic types of sensors to help them navigate. As the ones we tested worked their way through our corrals and one editor’s living room that we used as a control, we took stock of how efficiently they moved within the space.
Bump-and-Go: The vacuum heads in one direction and then changes course when it hits something. Many vacuums have programming that makes their paths more efficient. By bumping into a wall in a couple of places, they can verify the location of a wall and travel parallel to it, turning 180 degrees each time they hit an end wall.
Visual Simultaneous Localization and Mapping (vSLAM): This tracks multiple points in a room, in successive camera frames, to triangulate position. The Roomba s9+ employs this method to navigate. Over time it learns and becomes more efficient, where it may rely on points that remain constant and do not change.
Light Detection and Ranging (LiDAR): This consists of a laser that locates features by sending out pulses of light and measuring how long they take to return. LiDAR is especially useful for creating accurate maps and the most powerful navigation tool for robot vacuums.
How We Test
First, we set up three 8 by 8-foot pens, each with a different floor surface, where we could repeat cleaning scenarios for each model of robot vacuum. The floor of one corral was low- to medium-pile carpet; the second was bare, polished concrete; and the third was laminate flooring. We placed a floor mat and a wooden stool on the concrete as obstacles.
Then, in each corral, we dumped 5 grams of flour, 5 grams of sawdust, 15 grams of dried rice, and 15 grams of pinto beans to simulate various types of dirt and debris. (We should note that robot vacuums are maintenance cleaners, and these—for the sake of our stress test—were more material than you should expect a vac to pick up on a regular basis.) On the carpet, flour and sawdust are indicators of how well a vacuum’s brushes work at agitating dirt, making it easier to suck up. On the laminate, where sawdust and flour can get deep into seams, raw suction power is needed. And dried rice and pinto beans are a challenge on concrete and laminate, because a vacuum’s brushes could quickly scatter them.
Once we had the vacuums’ apps installed and paired, we set the models loose, timing and scrutinizing their work.
Water lift, a common method of quantifying suction, is relatively easy to measure. So easy, in fact, we built a simple water lift gauge to evaluate the strength of these robot vacs. Here’s how it works: We set up the vacuums so their suction could be applied, within a sealed system, to a tube with water in it. The stronger the suction, the higher the water is lifted.
Vacuum manufacturers tend to list vacuum strength in Pascals (Pa), which can be converted to water lift: 249.1 Pa equals 1 inch of water lift. You’ll notice our numbers are lower than manufacturers’ claims, and that’s to be expected because they don’t tell us where or how they’re measuring. They may take measurements at the impeller, or without going through the filter. It’s similar to how auto manufacturers list horsepower ratings on cars—as an engine spec. But the transmission, differential, and other systems all leech some of that horsepower, so what’s actually available at the rear wheels isn’t the same number listed in the specs. To be as close as possible to the actual amount of suction available under the vacuums, we removed the vacuum brushes and measured right through the housing above them, with the vacuum filter in place.
Roborock’s S7 MaxV replaces the S6 MaxV we tested previously. This new generation improves upon the previous model with more suction, faster processing speed to recognize obstacles, and a fully automated dock. Overall, the vacuuming speed remained similar to the S6. In the laminate corral, it vacuumed everything clean, leaving no trace of the debris we had strewn. It finished our carpeted corral in just over 9 minutes, with barely a trace of sawdust left behind. In our concrete area, with the rice we spread, the side brush scattered a few grains, which then got left behind. The small traces of debris missed, as with other models tested, is largely a byproduct of the sheer volume of detritus we used in testing.
Using both vSLAM and LiDAR to navigate, the S7 MaxV started with a diagonal zigzag path, presumably scanning the surroundings. It then quickly went around the perimeter of the corral, following that up with a series of out-and-back passes moving across the whole space. The vacuum then performed a second pass of the perimeter, ending with a final out-and-back series of sweeps, criss-crossing the previous passes, and then returning to the dock to top off the battery. The more times it vacuumed the same space, the more precise and efficient the robot’s path became.
While the mopping was sufficient to clean up the muddy footprints we left in testing, the automated features of the docking station are what really impressed us. Aside from the robot vacuum self-emptying into the dock, the onboard mopping reservoir is automatically refilled from the dock. Plus, the dock rinses and cleans the mop pad with an automated, spinning brush, directing waste-water to its own tank. So, with an empty dust bin and a tank full of clean water, the robot can run up to eight weeks unaided.
The S7 MaxV also pleased us during practical testing in our homes. In the Roborock app, we watched as the vacuum quickly created accurate maps of our rooms as it vacuumed in real-time. It also recognized and avoided sneakers, a power strip, and some wooden prop dog poop we threw down, leaving icons on the maps indicating what obstacles were and where they were encountered. In the app, we could view saved maps in 2D or 3D formats, edit maps, add virtual walls, add no-go and no-mop zones, as well as create room specific settings and cleaning schedules. Remarkably, you can also pull up live video from the robot while cleaning is in process, and even use it to place a video call to speak to whoever may be present in your home.
Yeedi’s Mop Station Pro is both a mop and a vacuum. With the large, rear dustbin attached, it works as a vacuum, running on a schedule you set to keep your house clean. When you want to mop, replace the dustbin with the mopping attachment that has two rotating, replaceable scrubbers. Unlike with may other mopping robots, which require you to fill an onboard tank, we were surprised that it filled automatically from a reservoir in the docking station. We discovered the dock also has a tank to hold dirty water from rinsing the cloth scrubbers after mopping.
During our mopping test, the Mop Station Pro performed well, scrubbing the laminate floor clean of our muddy footprints. The spinning scrubbing pads did kick some of the dried mud around, but our test pushes the limits of what we would expect the robot to mop up. While the floor was cleaned well, when we inspected the bottom of the robot, we needed to clean some mud off the wheels. In our own homes, we would normally grab a towel and quickly wipe up heavy mud before setting the robot mop loose to do its job. As a maintenance cleaner, the Mop Station Pro does a good job cleaning everyday types of drips, smudges, and marks on the floor.
As a vacuum, it performed well in our testing on all surfaces. We observed no trouble transitioning from hard floors to area rugs or floor mats, and the Mop Station Pro generally moved through the test areas in a methodical manner. It occasionally left a couple of grains of rice or a stray kidney bean, mostly due to the side brush flinging them across hard, smooth surfaces—something we regularly see on all vacuums with side brushes.
Paired with Yeedi’s app, setup, navigation, and mapping were straightforward. We were able to create no-go zones, set auto carpet detection to bump up suction for rugs, and schedule cleanings. For $800, you get a lot of features with the Mop Station Pro. And as the name implies, this vacuum is more biased than most toward mopping.
Shark’s AI Ultra robot vacuum is a hands-off maintenance cleaner with a HEPA filter, making it a great option for pet owners. The AI Ultra can run unaided up to 60 days because the docking station empties the robot at the end of every cleaning cycle. After each stage of testing, we examined the vacuum’s onboard dust bin after it self-emptied and found it clear of debris. We did note that there was always a small amount of dust residue sticking to the inside due to statice electricity. When the docking station does need emptying, we found it best to do it into a large garbage can in the garage or outside, because dust can become airborne, which sort of defeats the advantages of the unit’s onboard HEPA filter if done inside.
On our various test surfaces, the AI Ultra performed equally well on hard and carpeted surfaces. It occasionally left a few grains of rice or a stray kidney bean in our admittedly heavily deposited debris, but no more than the other models. Initially, the app directs you to run an exploration/mapping pass the first time you set it up in a new space. This allowed us to edit the map, adding no-go zones, carpet, and high traffic areas where it can increase suction. We could also assign areas to spot clean or have the AI Ultra make two passes instead of one.
We also tested the obstacle avoidance features when navigating our space. As long as the obstacle was as tall (taller than the vacuum), the robot would stop and maneuver around it. When we laid a roll of packing tape on the floor, the vacuum pushed it around. So, the vacuum will pick up individual Lego bricks, but it will avoid your child’s Lego castle. If these types of things are common in your house, you’ll want to set up no-go zones for children’s play areas.
The AI Ultra will also function as a robotic mop, when you swap out its onboard dust bin with the water tanks and mop pad. Our muddy footprints were a lot to ask of it, but the AI Ultra did a very good job of wiping the area clean, with just a couple of crumbs of dirt left behind. We did have to thoroughly clean the mop pad after the test, as it was caked with mud—but our test was beyond a normal mopping scenario.
The s9+ clocked the fastest time to finish the carpeted area, leaving just a trace of sawdust after 9 minutes. In the concrete area, it took more time navigating around the wooden stool but was delicate in doing so, cleaning around the legs thoroughly. Like the others, it did leave some rice at the edge of the floor mat. In the laminate corral, we couldn’t find any debris except for a dusting of flour in one spot. Keep in mind that the s9+ has flexible ridges and paddles instead of bristles. This appeared to be a little less effective with very fine particles on the smooth, hard surface.
We had the most difficulty discerning what the s9+ was actually doing. What initially appeared to be a sort of random path, we realized, was a series of location exercises using vSLAM. It first moved left and right, bumping both sides in multiple places, and then it worked toward the corners, looping into them in a wide curve before moving tight in and backing out. Then it performed a trip around the perimeter, and then into the corners again. After all that, it moved to make left and right passes to cover the middle area. Finally, it vacuumed left and right of the dock.
This Roomba has a slightly different design than most other robot vacuums, in that it has a flat front with squared-off corners. We found two advantages of this, the first being that it got into corners better, and the second that it allows for wider brushes—about 30 percent so.
The G30 Edge is the least expensive model from our test, but you wouldn’t know that by how it cleaned our test corrals. It was slightly slower than the others, sure, yet it effectively and methodically vacuumed each area. It left barely a trace of debris on both the carpet and laminate floor and missed some rice transitioning from the floor to the mat in the concrete area. We did notice the G30 Edge bump things slightly harder than the other vacuums— not surprising, since its navigation is an enhanced version of bump-and-go.
Despite having that lower-tech nav system, this Eufy seemed remarkably efficient. Leaving the dock, it immediately started a straight out-and-back pattern, moving left to the wall with each pass. It then bumped the wall in a couple of places, confirming its location, before returning to the center. It continued out and back to the right wall, where it started around the perimeter to the lower left corner. It then began left and right passes from the front to the back, followed by the perimeter, ending with the lower right corner.
Testing in our homes, we used the app to monitor progress in real time. While we could view completed vacuuming and maps in the cleaning history, we couldn’t save or edit them. To set no-go zones, we laid down the included magnetic strip to create physical barriers. While this worked well, we did need to remember to put the strip in place every time we had the vacuum scheduled to clean.
Brad Ford has spent most of his life using tools to fix, build, or make things. Growing up he worked on a farm, where he learned to weld, repair, and paint equipment. From the farm he went to work at a classic car dealer, repairing and servicing Rolls Royces, Bentleys, and Jaguars. Today, when he’s not testing tools or writing for Popular Mechanics, he’s busy keeping up with the projects at his old farmhouse in eastern Pennsylvania.