It’s tricky to make a robot vacuum navigate through your home without getting stuck or lost. Bruno Hexsel, a former software engineer for Neato, told us that he spent a huge chunk of his time at the company working on algorithms to help robots get unstuck from common hazards.
It’s not so different from a self-driving car. Yeah, the stakes are much lower for a 9-pound robot blooping around your living room than they are for a 3,500-pound car hurtling down a public highway. But they both need to navigate their environments on the fly, with an unpredictable set of obstacles and hazards, in unique combinations, that can change constantly.
The challenges are similar enough that several former robot-vacuum engineers now work on building autonomous-vehicle nav systems—including Hexsel, whose current employer is in stealth mode. Duane Gilbert, formerly of iRobot, works at Nio (as of June 2018), a startup working on self-driving electric cars.
Gilbert told us that it’s arguably harder to get a robot vacuum to navigate reliably, because your home actually has more variables than a road. Floors don’t have traffic lines or mile markers or light poles to orient the bot, and the traction on different types of rugs is much trickier than on just asphalt. And you can put more sensors—and of a higher quality—in a $40,000 car than you can in a $700 robot.
So what does a robot vacuum need to do in order to navigate a real home?
Don’t get stuck
Over many years of using robot vacuums, we’ve found that the robots that keep moving are the ones that keep your floors the tidiest. So the most important challenge for any bot is to avoid getting stuck or otherwise quitting mid-cycle. Sophisticated mapping and powerful suction don’t mean anything if the bot stops running 10 minutes after you leave for work.
The most common bot-trapping hazards include power cords, charging cables, stray laundry (especially socks), dangling curtains or bed sheets, rug edges, floor registers, tall thresholds (some bots can’t get over them), and furniture with tube-shaped or extra-wide supports that lie across the floor. Higher-end models can have trouble with dim lighting (if they have a camera) and chromed furniture (if they use lidar). Black rugs (or other dark, nonreflective flooring) look like a bottomless pit to the anti-drop sensors, so many bots won’t clean them. Some homes have more of these traps than others; most homes have at least a few.
Great bots can deal with almost all of those hazards, while others are helpless around most of them. So many factors affect their performance around obstacles: The number and placement of bump sensors, the tuning of the drop sensors, the way the brushes spin, how the robot senses tangles or jams, the size of the wheels, the spring tension and pivot placement in the suspension—not to mention the algorithms that turn the input from the sensors into robot movements. Spending more doesn’t guarantee that you’ll get a nimble, trap-escaping robot. Some of the most expensive models are also the most likely to get stuck, and vice versa.
Don’t get lost
The second-biggest challenge is to clean as much of the house as possible without missing many patches. There are a couple ways to accomplish this.
The simpler, cheaper way is a semi-random, bump-and-run navigation style, where the robot essentially bumbles around the house until its battery runs out, not following any particular path, making semi-random turns whenever it comes across an obstacle and can’t keep driving in a straight line.
As dumb as they can look, we’ve found that bump-and-run models often pick up more total debris than “smarter” robots because they are so persistent, making two or three passes in a single session (as long as the area isn’t too big). Sure, the bump-and-run bots might miss a patch or even an entire room one day, but if you run it at least a few days per week, it should end up getting around to everything often enough that your floors always feel totally tidy. “I once watched [my bot] vacuum the same 5-by-8 area for a solid 30 minutes,” said Wirecutter staff member Gustave Gerhardt, “but mostly its randomness is effective.”
Bump-and-run bots aren’t the best fit for everyone. Some people feel stressed out from watching its random nav system miss a big tuft of cat hair over and over. Larger homes increase the chances that these bots will miss patches (1,200 square feet per level is a practical limit). And a choppy floor plan with narrow doorways increases the chance that the bot will spend too much time cleaning one particular section while it fumbles for an exit.
If you’re willing to pay more, higher-end mapping bots figure out the layout of your home, and follow a more orderly, grid-like cleaning path so that they don’t miss anything or overclean any room. They can also reliably return to their docks, recharge their batteries, then pick up cleaning where they left off if they hadn’t finished. Essentially, they should clean an entire level of your house, every single session, no matter what your floor plan looks like. That’s useful for bigger homes, where bump-and-run bots might get lost. Plus, life is chaotic enough, so if you’ll get some peace of mind from a robot that drives in a predictable pattern and consistently returns to its dock, treat yourself.
Mapping systems come in a few varieties:
Some mapping bots feel their way around your house sort of like a bump-and-run robot, except they follow a grid-like pattern and typically use a ceiling-facing camera (among other tricks) to track their position. You could call them bump-and-track bots. Essentially, they draw a map based on what they’ve already encountered.
Other mapping bots use a rangefinder to pre-map each section of your house. That is, they scan for obstacles and draw most of their map before they start cleaning any particular area. Most of these models rely on lasers (lidar), though we’ve seen some that rely on a camera or laser-and-camera combo.
Our take: Mapping bots are not strictly better than bump-and-run bots, even though they can cost two, three, even four times as much. They aren’t necessarily better at avoiding or escaping traps. The extra complexity in their nav systems leads many of them to get confused more easily, so that they’re more likely to quit cleaning in the middle of a session. Plus, mapping bots with weak suction pick up less debris than bump-and-run bots, because they make only a single pass in any cleaning session. Just because a mapping bot looks smarter doesn’t mean it’ll be better at cleaning your house. So before you pay extra, consider a bump-and-run bot first.
Suck up the obvious junk
Cleaning performance obviously matters, though it’s not as important as most people think. As long as your floors look tidy from eye level, and you don’t feel like any crumbs or hair are sticking to your feet when you walk around, the bot is doing its job. If a bot sucks some extra-fine dust out of your rugs, that’s a nice bonus, but it’s not the purpose of the product. Even the strongest, most expensive robot has a fraction of the cleaning power of an $80 upright vacuum. You can’t rely on a robot to deep-clean your carpets, so we don’t think it’s worth fretting over which bots have the absolute-strongest cleaning power.
At a minimum, robot vacuums need a brush roll and preferably at least one side brush to clean effectively.
As with any kind of vacuum cleaner, suction and brush-roll action both affect cleaning performance. For bump-and-run robots, battery life is also a factor—more time on the floor means more opportunities for cleaning (though it’s not very important for mapping robots, because they can recharge themselves mid-session and pick up where they left off).
If you have more of one key specification, you can get away with less of the others. Some bots have strong suction but less brush action; for others, the opposite. Many have fairly weak cleaning power, but enough battery life to make multiple passes. (Some models have adjustable suction settings so you can choose the balance of battery life and suction that suits your layout best, though they don’t always make a noticeable difference.) In our experience, several of these approaches are valid, able to pick up a lot of debris.
That said, if you have thicker rugs (fibers ½ inch to ¾ inch) and long-haired pets, you should pay extra for a robot with a little extra cleaning power than the cheaper bots. No matter how many passes the low-powered ones make, they just won’t pick up enough fuzz to make most people happy in these situations.
What else matters
In descending order of importance, roughly:
Quieter robots are easier to tolerate if you have to be at home while they’re running. We measured the Eufy 11S at 54 dBC (measured from 10 feet away), which is really quiet. 60 dBC was more typical in our tests. Several models run near 65 dBC when they’re on bare floors, which is loud enough to be annoying.
Robots that are relatively short and lightweight have an advantage over taller, heavier bots, because they can get under more furniture and tend to be better at driving over thresholds and bare-floor-to-carpet transitions. Anything under 3 inches is very short. Apart from height, we’ve found that the shape of the bot isn’t super important: Yeah, D-shaped bots fit into corners better than round bots do, but round bots have side brushes that mostly accomplish the same thing, as long as you’re not purposely sprinkling powder up against your baseboards to test their abilities.
Good bots should have replacement parts that are easy to find—at least filters, but preferably brushes, batteries, wheels, and even the transmission. Knockoff filters are okay, but moving parts or electronics should be bought from the manufacturer, so that you don’t risk damaging the robot. It’s also helpful when the bots are easy to take apart for cleanings or repairs.
Boundary markers can be useful if you don’t want your bot crossing into a certain room, plowing into your pet’s water bowl, or hurtling into an area where it always gets stuck. We like the Roomba virtual walls the best because they’re reliable and unobtrusive. Several bots now let you draw boundaries using the smartphone app—a great idea with room for improvement. Other bots rely on ugly magnetic strips, laid across your floor, to turn the bot away. Many cheap bots have no boundary markers at all.
Wi-Fi connectivity can be useful when it works correctly. It lets you add your bot to your home wireless network so that you can control it from a smartphone app, or sometimes through a voice assistant (like Alexa or Google Home). The obvious use is that you can start or stop the bot while you’re away from home. Apps also make it easy to set a cleaning schedule, keep track of when you need to replace parts, or adjust certain cleaning settings. (Bots without Wi-Fi usually come with a physical remote control instead, which is still handy.)
Like anything involving wireless networks, nobody can guarantee that your robot’s Wi-Fi will work smoothly. And this type of troubleshooting can make you shake with rage. Plus, all robots have the ability to collect at least some data about your floor plan (the mapping bots can collect quite a bit), and if they connect to Wi-Fi, there’s always some chance that the data falls into unwanted hands. For example, a security firm figured out how to look directly through the camera on certain LG robot vacuums. And a minor panic broke out after a Reuters article misreported that iRobot had plans to sell owner data to third parties. We haven’t looked closely enough at any brand’s terms of service to know what they are doing (or could do) with your data. But if you’re annoyed or creeped out by the whole thing, you can either buy a non-connected bot or just never set up the Wi-Fi—these bots all still clean your house automatically even if you don’t connect them to the Internet.
Most bots have a few different cleaning modes beyond the main automatic whole-house setting. In our experience, most people never use any of them, but you might find a spot-cleaning mode (very common) or the manual steering option (less common) convenient for taking care of small, contained messes.
The size of the dustbin is mostly inconsequential. If you have pets, yes, the bin might fill up quickly the first few times you run a good robot, as it works through the backlog of hair that you didn’t know had built up around your house. But after a week of regular cleanings, give or take, it should start to come back with less and less hair. The differences in bin size from model to model aren’t that substantial, anyway. So if your pack of shaggy dogs stuffs up one robot, it’ll probably stuff up every robot (unless it can empty itself mid-session, like the Roomba i7+).
We do read lots of owner reviews for any robot that we consider recommending, though we don’t pay much attention to the average scores anymore. Reviews are absolutely useful for learning about poor reliability or customer service (most bots have a one-year warranty, for what it’s worth), or navigational problems that we didn’t find on our own. It also helps us understand what our readers might want to know and what’s actually important to them. But review manipulation has become common, so we don’t trust average owner ratings as much as we used to.