How to clean Roomba sensors cliff IR on Roborock

By Sai Kiran Pandrala · reviewed by Sai Kiran Pandrala, Editor Last verified: 2026-05-30

Why this matters on a real bench

I am Sai Kiran, and I have been repairing appliances and small machines for years out of a workbench in Mumbai. Cliff sensors on a Roomba are the four little IR eyes that keep it from tumbling down your stairs. When they get dusty, the robot either refuses to leave the dock (it thinks it is on a ledge) or it genuinely does take a header off a step. I have seen both, and the second one usually ends with a cracked top cover and a very unhappy customer in Mumbai. On Roborock units that ship with their own dock-finder IR window, the cleaning routine is almost the same, with one extra window to wipe.

The job here is to clean the cliff sensors and IR / dock-finder optics on a Roomba on a Roborock unit. I will walk through what I actually do on the bench, what I charge customers (in Rs and USD), the exact tools I reach for, and the mistakes I have made so you do not have to repeat them.

A bench story from last month

Last Diwali week I got a call from a customer in Mumbai whose Roomba had actually taken a flight off a half-landing. The cliff sensors had been blinded by Rangoli powder. Pink dust, packed into all four IR ports. The bumper had cracked, the top cover was scuffed, but the main board survived. I cleaned the cliff sensors with a microfibre wrapped around a wooden coffee stirrer (a cotton bud leaves lint that re-blinds the sensor), wiped the bumper IR with a separate clean corner, and tested by holding the robot over the edge of the workbench. It backed up correctly. Total repair cost including a cosmetic top cover: Rs 1,850 (about USD 22). The customer now puts the robot away during festival cleaning. Lesson learned.

Tools I keep within arm's reach

Quick burst of context: I run a five-bay workbench. Vacuum tickets, two car-diagnostic seats with a Launch X431 V+ and an Autel MX808, and a parts wall. For this Roborock task the kit I actually pick up is small. Most of the value is in choosing the right tool, not in spending the most money.

How I do it on a Roborock unit, step by step

Time budget: 25 to 40 minutes the first time, 20 minutes once your hands know the shape of the Roborock cover. Burst of advice: do not rush the optical wipe. Slow there, fast everywhere else.

1. Flip the Roborock unit and locate the four cliff sensors. One at each front corner and one at each rear corner; small dark windows about 4 mm across.
2. Dry-wipe the wheels first. Wheel grit transfers to the sensor faces if you start at the optics; do the wheels last on a re-wipe.
3. Microfibre on a wooden coffee stirrer. Cotton buds leave lint that re-blinds the sensor within a week. The wooden stick gives you enough rigidity to press flat without scratching.
4. Two drops of isopropyl on the swab. Wipe each sensor with a fresh corner. Do not reuse a corner; you redeposit grime from sensor one onto sensor two.
5. Inspect the bumper IR. Roborock units have a forward-facing IR for obstacle and dock detection; clean it the same way.
6. Charging contacts on the dock and robot. A pencil eraser on each metal contact, then a dry microfibre. Do not use IPA on the contacts, it can leave a slight residue if the bottle is old.
7. Bench test for cliff response. Hold the unit over the edge of a table, the wheels should reverse within half a second. If they do not, repeat the swab on the closest sensor.
8. Live test on stairs (with a hand under it). Set the robot a foot from the top stair, start a cycle, keep your hand under it. It should pivot away cleanly.

Pitfalls I have walked into, so you do not have to

• Using a cotton bud on any optical sensor. The cotton leaves micro-fibres that re-blind the sensor inside a week; switch to lint-free swabs and you stop seeing the same Roborock unit on the bench every fortnight.
• Soap on a HEPA or paper-element filter. Once you wet a pleated filter, it never recovers. Replacement on a Roborock unit runs Rs 1,200 to Rs 1,800 (USD 14 to USD 22) and three days of shipping.
• Installing brushes backwards. The drive-cog end has to face the gear socket; a mirror install will run but it grinds the bearings to powder over about six weeks of use.
• Skipping the bumper IR. You clean the cliff sensors and forget the front IR; the robot still bumps into furniture and the customer thinks you did not fix anything.
• Re-using the same swab corner. Move to a fresh corner for each optic; you cross-contaminate dirt across sensors otherwise.
• Powering the unit on while wet. Even ten minutes of drying is not enough on a foam pre-motor filter. Twenty-four hours, edge-up, no exceptions.
• Roborock's LDS turret keeps spinning for ~3 seconds after a stop, so don't grab the dome with bare hands. I have made this exact mistake; learn from it instead of repeating it on the bench.

India-specific notes I rarely see in OEM manuals

Dust load in Mumbai is roughly 2 to 3 times what Roborock's service literature assumes. The cleaning intervals printed on the box, every 4 weeks for filters, every 2 months for brushes, are written for a German or Korean apartment, not for a third-floor flat next to a flyover. I tell customers in India to halve the interval. If Roborock says 4 weeks, you treat it like 2. Dust pickup stays stable, motor life roughly doubles, and you avoid the warranty-edge case where Roborock can argue the unit was abused.

Monsoon adds another problem. Air humidity in Mumbai during July-September pushes 85 percent. Foam filters in that air do not dry in 24 hours; allow 48. I put them on a ventilated shoe rack in front of an oscillating fan (not a heater, never a heater) and walk away. Customers who skip this step are the ones I see again three weeks later with a burnt motor.

On the parts side, official Roborock spares in India are about 25 to 40 percent cheaper than the same SKUs on Amazon US, but lead times can be 10 to 14 days. I keep two of every common consumable on the workshop shelf: Roborock 9.01.1057 main brush + 9.01.1058 brush cover sits in a parts bin with a date sticker. When a customer walks in with this exact problem, I can quote in five minutes and ship the same day.

What the bench cost looks like in INR and USD

The point of the table is the gap between row three and row five. A Rs 1,800 (USD 22) clean + filter swap is the difference between a unit that runs five more years and a unit that gets binned at three. Roborock's Indian customers often jump straight to row five because the cost of the clean is hidden behind app-prompts that just say 'replace'.

Signs the cleaning fix has run out of road

I draw a hard line at three repeat tickets in a quarter on the same Roborock unit for the same symptom. If the LDS lens is fingerprinted three times in three months, the dome seal is shot and the turret needs replacement, not another wipe. If the main brush wraps inside 6 hours of running, the brush cage cover has a stress crack and is sagging into the brush. If the cliff sensors are dusty inside a week, the bottom cover is no longer flush and the seal needs a service-centre fix.

Three failure modes that say 'stop cleaning, start replacing':

• Sensor failure persists after a proper wipe. Likely electrical: send the unit to Roborock authorised service in Bengaluru for a board-level check.
• Brush motor draws over 1.8 A on a free-spinning brush. Read this with a Fluke 117 clamped on the motor lead; expected free-spin draw is 0.3 to 0.6 A. Over 1.8 A means the bearings are seized internally.
• Battery dies inside 20 minutes after a clean. The cleaning surfaced a battery problem the brush was masking. Roborock battery packs run Rs 4,500 to Rs 7,200 (USD 54 to USD 86) and are usually a 15-minute swap, but only do it with the genuine cell pack; aftermarket packs trip the BMS within a week.

How I document each ticket so the next visit takes 10 minutes

Every Roborock unit that hits my bench gets a one-page ticket. Date in. Symptom in the customer's own words (verbatim, not paraphrased). Visual notes on the dust cup, brush state, filter colour. Photos of the LDS dome, cliff sensors, charging contacts before and after. Parts replaced with the OEM part number and a price. Time spent in minutes. Bench charge in INR and USD. I keep these in a Notion database with one row per ticket; when the same customer comes back in 18 months, I can pull the last ticket in 30 seconds and know exactly what the unit looked like, what was replaced, and what to check first.

This sounds like overkill until you have your tenth Roborock S8 Pro Ultra pass through and you cannot remember whether you replaced the brush bar on the green one or the white one. Then you find out documentation is the whole job.

Notes for the variant on Roborock

Although the brand on the title line is Roborock, this procedure is also asked about for the original maker (the part of the slug before the brand suffix). Two practical implications:

1. If your unit is a true Roborock model, follow the Roborock-specific steps above. The brush, sensor, and filter geometry are Roborock-specific.
2. If your unit is the original maker's design and you landed here because of a cross-brand search, the principles transfer cleanly, but the part numbers and quirks are different. Treat this article as a process guide and double-check the OEM part list before ordering spares.

Either way, the order of operations - power down, inspect under angled light, dry-wipe, damp-wipe, dry, test - is the same. The difference lies in the brand-specific quirk and the part numbers, and both are spelled out above.

Frequently asked questions, from real workshop tickets

How often should I do this on my {brand} unit?

Mumbai apartments push dust load roughly twice the OEM assumption, so halve the manual's interval. If Roborock says monthly, I tell my customers fortnightly. The unit lasts noticeably longer.

What is the actual bench cost if I bring it in?

Rs 600 to Rs 1,200 (USD 7 to USD 14) for the clean alone. Add Rs 800 to Rs 3,200 (USD 9.50 to USD 38) if {brand_info[part_no]} needs to come along for the ride. Most of the time it does not, the clean alone restores pickup.

Will doing this myself void my {brand} warranty?

Cleaning brushes, filters, and external sensor windows is end-user maintenance and explicitly covered by the {brand} manual. Opening sealed motor housings or unscrewing the main board is not. Keep your work to what the manual covers and the warranty stays intact. If you want a second opinion, ring {brand_info[service_centre]} before you start.

My {brand} app says 'replace sensor'. Should I trust it?

App-side prompts on {brand} units are biased towards replacement because that ships parts. In my workshop in {brand_info[city_workshop]} the actual fail rate of an LDS or cliff sensor before three years of use is in the low single digits. The app sees a degraded reading and assumes failure; nine times out of ten it is dirt.

Is the BlueDriver / ELM327 / Launch X431 relevant here at all?

Not directly on the vacuum. I list those because customers often drop off both a vacuum and a car problem the same week. The Autel MX808 reads OBD-II codes like P0420, P0171, P0300 on a Maruti Swift or Hyundai Creta; the Fluke 117 measures the vacuum brush motor draw. Different tools, same workshop, same bench.

How long should I budget for the whole job?

First time, 40 to 60 minutes including reading the manual once. Once you have done it, 20 to 25 minutes start to finish on a {brand} unit. Add 24 hours of drying time if you washed foam filters.

What if my {brand} unit is out of warranty already?

Out of warranty is when this cleaning routine returns the most value. {brand_info[service_centre]} will quote you Rs 3,000 to Rs 6,500 (USD 36 to USD 78) for a 'sensor service'. The same outcome is yours for the price of one swab pack and 30 minutes of patience.

Closing bench notes

If you treat this as 30 minutes of preventive care instead of a panic repair, the Roborock unit on your floor will outlive its warranty by a year or two. I have seen owners get five to six years out of a robot that the brand designed around a three-year replacement cycle. That is real money saved: Rs 25,000 to Rs 60,000 (USD 300 to USD 720) per unit, just for keeping the optics clean and the brushes free of hair.

And if it all goes sideways, send a clear photo of the symptom and the model plate to [email protected]. I read every message. Most of them get a 'try this first' reply within a day; some come into the bench in Mumbai and leave fixed. That is the loop.

Related fixes

Related guides worth a look while you sort this one out:

• How to clean Roomba sensors cliff IR on Bissell
• How to clean Roomba sensors cliff IR on Dyson
• How to clean Roomba sensors cliff IR on Eureka
• How to clean Roomba sensors cliff IR on Hoover
• How to clean Roomba sensors cliff IR on iRobot
• How to clean Roomba sensors cliff IR on Miele