Key takeaways
- GPS signals cannot reach a submerged robot, so nothing underwater navigates by satellite.
- Robots estimate position by dead reckoning: a gyroscope and IMU track every turn while software counts distance.
- Small sensor errors accumulate as drift, and wall contact is how the robot zeroes its estimate.
- Random bounce, gyro rows, and sensor fusion are the three tiers; they differ in cycle time and consistency, not scrubbing power.
- Missed strips usually trace to start position, obstacles, or slipping tracks, all fixable without buying a new robot.
Here is the question every owner eventually asks me: how does this thing know where it has been, under six feet of water, where no GPS signal survives? The honest answer is that it does not know. It estimates.
Every pool robot, from a $300 bouncer to a $2,000 flagship, navigates blind. The good ones just estimate carefully, check the guess against the walls, and start fresh before the errors grow teeth.
Once you see how the trick works, your robot’s odd habits, the mid-floor pauses, the wall hugging, the potato-shaped map, all start making sense.
Why there is no GPS at the bottom of your pool
Satellite signals arrive faint even in open air, and water absorbs radio energy aggressively. A foot or two of pool water swallows GPS, Wi-Fi, and Bluetooth alike, which is why the app goes quiet the moment the robot submerges.
That left engineers two options: run positioning down a wire from the surface, or teach the robot to track itself with no outside help at all. Everyone chose the second, borrowing a method sailors trusted for centuries before satellites existed.
Dead reckoning: counting steps across a dark room
The method is dead reckoning: if you know where you started, which way you face, and how far you have traveled, you always know roughly where you are.
You could cross a dark bedroom that way and do fine for ten paces. The robot plays the same game with better instruments: a gyroscope reports every degree of turn, an inertial measurement unit adds acceleration and tilt, and the drive system logs how far the tracks and wheels have rolled.
Stack those readings many times a second and the robot carries a running guess of its own position, good enough to lay parallel rows across a floor it cannot see.
Drift: why the estimate slowly rots
Every sensor reading carries a speck of error, and dead reckoning has no outside referee to cancel error out. It accumulates. Engineers call the result drift, and it is why a robot trusting its gyro alone would end a 2-hour cycle scrubbing a corner that does not exist.
Slipping makes drift worse. A track that spins on a slick fiberglass slope logs distance the robot never covered, which is how worn tracks and wheels quietly corrupt navigation long before they stop the machine.
Watch a cycle closely and you will catch the countermeasure: the robot halts mid-floor for a second or two, apparently thinking. It is re-zeroing the gyro, a small calibration pause that keeps the guess honest.
Walls: the reset button that never lies
The real correction is physical. Walls do not move, so every arrival at one hands the robot a fact it can trust, and the software leans on those facts hard.
Plenty of models open a cycle with a wall-follow lap, sketching the pool’s outline before filling in the middle. Each later wall touch then snaps the drifting estimate back onto that outline.
Wall climbing earns its keep here too. A run up to the waterline gives the robot a clean vertical reference and a known restart point for the next stretch of floor, on top of the scrubbing it was built for.
The three navigation tiers you can actually buy
Manufacturers package all of this into three recognizable tiers, and the differences show from a deck chair.
| Tier | How it steers | Typical price | What you see in the pool |
|---|---|---|---|
| Random bounce | No position estimate, timed turns off walls | $200-450 | Chaotic paths, 2.5-3 hour cycles, everything covered eventually |
| Gyro rows | Dead reckoning corrected by wall contact | $500-1,000 | Parallel lanes, 1.5-2 hour cycles, the odd missed strip |
| Sensor fusion | IMU plus ultrasonic or camera input | $1,000-2,500 | Deliberate coverage, obstacle detours, steadier waterline work |
The middle tier is where navigation stops being luck. A Dolphin Premier or an Aiper Scuba S1 laying tidy rows finishes a plain rectangle in about half the time a random-walk machine needs, with the same brushes doing the scrubbing.
The top tier adds eyes. Ultrasonic pingers and cameras spot walls and obstacles before contact, which mostly pays off in freeform shapes and cluttered pools rather than open rectangles.
What the app map actually is
The map screen is a drawing of the robot’s estimates, not a survey of your pool. It exists to reassure you, and it redraws a little differently every run because the estimates differ every run.
It drew my kidney pool as a lumpy potato, but the floor is spotless. I stopped grading the art.
That owner has it right. A few flagships now keep maps between cycles and quote coverage percentages; read those numbers as a mood, not a measurement.
A ten-minute test you can run from a deck chair
Last June I watched an Aiper Scuba S1 work a kidney-shaped gunite pool as an audit for its owner. From the deck you could see the opening perimeter lap, then rows laid at a slight angle, then a second pass rotated across the first. One missed crescent near the steps vanished the next run, once we started the robot from the opposite corner.
Your pool will show the same signatures if navigation is healthy: an orderly opening lap, row spacing with a little overlap, confident straight lines across the deep end.
Wandering arcs, repeated scrubbing of one zone, or a robot that never settles into rows means the estimate is failing, usually for a reason you can fix: a slick surface, glazed tracks, or an awkward start point.
When the rows go wrong
Missed strips and lopsided coverage have a short cause list, and a dead robot is not on it.
Pro tip
Rotate the drop-in point. Dead reckoning cycles are anchored to where they begin, so three runs from three different spots cover a freeform pool far more evenly than three identical starts ever will.
Beyond start position, the fixes are mundane. Pull the ladder and toys before a cycle, keep the water above roughly 60F so the drive keeps its grip, and replace tracks the season they start slipping.
If coverage collapsed suddenly rather than faded gradually, check the app for a pending firmware update before suspecting hardware. Navigation tuning arrives over the air more often than owners expect.
Does smarter steering mean cleaner water?
In a 12×24 rectangle, barely. Random coverage gets there given time, and time is cheap when the robot runs on a cleaning schedule while you are at work.
The payoff lands in large pools, freeform shapes, and yards where debris returns daily: fewer missed strips, shorter cycles, less wear per week of clear water. That is also the honest ceiling. Navigation buys coverage and efficiency, while suction, brushes, and filter media still decide how clean each covered inch actually gets.
Reading your robot from the deck
Navigation underwater is estimation plus humility: count your steps, trust the walls, start fresh often. Once that clicks, the pauses look like calibration, the wall hugging looks like map-making, and the potato map looks like honest work.
Spend ten minutes actually watching a cycle this week. You will learn more about your robot’s coverage pattern from a deck chair than from any spec sheet ever printed.
Owners working through this often end up reading Pool Robot App Keeps Disconnecting: Staying Paired as well.
If your situation looks different, Robotic Pool Cleaner Buying Guide: Match Robot to Pool is a sensible next read.
Frequently asked questions
Can I set no-go zones or virtual walls for a pool robot?
Almost never, as of the current model years. Unlike vacuum robots, most pool robots cannot hold a persistent map accurate enough to honor drawn boundaries, so apps rarely offer them. A few camera-equipped flagships are inching that way with obstacle avoidance. The practical workaround is physical: remove ladders and toys before cycles, and choose floor-only mode when you want the robot kept off walls and steps.
Does the robot remember my pool from one run to the next?
Most models start every cycle from scratch: new dead-reckoning estimate, new wall corrections, no memory of yesterday. That is by design, since drift makes an old map less trustworthy than a fresh perimeter lap. Some premium machines do persist a learned outline and show cumulative coverage in the app, but even those re-verify against the walls constantly rather than trusting the stored picture.
Does cloudy or murky water confuse navigation?
Not for gyro and IMU machines, which never look at the water at all; they steer by motion sensing and wall contact and work identically in a green pool or a clear one. Camera-assisted models are the exception, since turbidity blinds the lens and they fall back on inertial guidance. Cloudy water is harder on filters than on navigation, so expect basket rinses, not lost robots.
Why does my robot climb the same wall again and again?
Usually because that wall sits where its coverage logic schedules climbs, and the estimate resets each time it returns to the floor nearby. Once per area is normal; obsessive repeats suggest the robot is losing its position estimate and using the wall to re-anchor. Try a different drop-in point, check the tracks for slip, and confirm the surface near that wall is not freshly slick with algaecide or new sealant.
Do ladders and main drains throw off the robot's path?
They deflect it, and dead reckoning absorbs small deflections fine because the next wall contact corrects the estimate. The bigger cost is the shadow: a fixed ladder can leave a small unswept patch behind it every single run from the same start point. Rotating where you drop the robot in moves that shadow around the pool, which is the cheapest coverage upgrade there is.
Krisha Gupta
Krisha Gupta is a home-technology writer who spends most of her weekends knee-deep in pools, testing the latest robotic cleaners so readers don't have to. Her reviews are hands-on and jargon-free, focused on what a robot actually does when you leave it running unsupervised for eight hours. When she's not comparing brush patterns or waterline scrubbing, she's writing plain-English maintenance guides for pool owners who'd rather swim than tinker.
