Robot vacuums have undergone significant improvements since the first Roomba hit the market. Over the past several years, we’ve seen advances in cleaning, navigation, mopping, self-emptying base stations, and even the ability to climb stairs and dodge obstacles.
But there are still limitations for these autonomous floor cleaners. Object recognition and evasion remain a challenge, and none of the 24 robot vacuums we tested avoided all six obstacles.
Our exclusive lab data shows which robot vacuums best combine cleaning power with object avoidance, coming closest to the perfect machine. While none of them are flawless, the standout was the Mova V50 Ultra, which posted the highest average score and strong pickup across all flooring types. This earns it our lab award for the best pickup performance, and it joins the ranks of our other award winner, the Dreame X50 Ultra, which boasts the best cleaning coverage.
Don’t miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.
Key takeaways:
- Consumers must sacrifice maximum cleaning power for reliable object avoidance. Our best cleaning performers, such as the lab award winner, the Mova V50 (with the highest overall average pickup at 65.14%), only avoided three out of six obstacles, while the Roomba 205, with the highest hardwood score, avoided only one.
- None of the 24 robot vacuums successfully avoided all six objects we tested, and most of our best cleaning performers fell in the middle of the pack, despite nearly every company’s marketing claims about superior obstacle avoidance.
- Of the four robots with the best object avoidance, only one, the Yeedi M14 Plus, scores above 50% in average pickup.
- Seven robot vacuums failed to avoid any obstacles at all: the 3i S10 Ultra, Narwal Flow, Eureka J15, 3i P10 Ultra, Monsga MS1 Max, Roomba 705 Max and SwitchBot S20 (the non-autofill version).
- Several robot vacuums experienced critical obstacle-avoidance failures, running over simulated pet waste, making them unsuitable for pet owners.
How we measured cleaning performance and object avoidance
You can see a comprehensive breakdown of how we test robot vacuums, but below are the two key metrics we used for this lab study:
- Cleaning performance: We measure this by the average pickup percentage, which represents the average amount of sand collected across hardwood, low-pile and midpile carpets. A higher percentage indicates more effective cleaning.
- Object avoidance: This is measured by the number of common household items avoided out of a total of six. Items included simulated pet waste, a lamp cord, a pet toy and a sock placed at different angles of approach to the robot vacuum.
Robot vacuum average pickup vs. object avoidance: A tough trade-off
During our testing, we discovered something surprising — you can have a robot vacuum with great cleaning performance or a robot vacuum with great object avoidance, but it was rare for a single model to be proficient at both.
Out of the 24 robot vacuums we tested, only nine of them offer a decent balance of cleaning and object avoidance, which we define as at least three objects avoided and an average pickup of at least 40% and higher. Four were able to avoid five out of the six objects, and of those, only one, the Yeedi M14 Plus, was able to score above 50% average pickup.
Powerful cleaners have worse object avoidance
Key takeaways:
- The overall top performer in cleaning power, the Mova V50 Ultra Complete, achieved the highest average pickup at 65.14% but only avoided three out of six obstacles.
- The Dreame D10S Ultra, which achieved an average pickup of 62.32%, only avoided four out of six items.
- The Dreame X50 Ultra, the second-highest average pickup performer at 58.64%, also avoided only three out of six obstacles.
- The Roomba 205, which achieved the highest hardwood score (99.27%), performed poorly in avoidance, successfully avoiding only one out of six items and failing to avoid the sock, pet toy and all three types of simulated pet waste in its first trial.
The units with the highest average pickup score often perform poorly in avoiding common household hazards. For example, the Mova V50, which achieved the highest overall average pickup at 65.14%, only avoided three out of six obstacles, even with its intelligent avoidance setting turned on. The Roomba 205 DustCompactor (the highest hardwood pickup at 99.27%) avoided only one out of six items.
As mentioned, the Mova V50 is the champion for cleaning, earning the highest overall average pickup at 65.14%. This high score is particularly notable because the Mova V50 achieved an unprecedented 47.54% pickup on mid-pile carpet, the highest score out of 24 robot vacuums. However, this raw power comes with navigational blind spots: Despite having its “intelligent obstacle avoidance setting” turned on, the Mova V50 still failed to avoid three out of six obstacles.
Similarly, the Roomba 205 DustCompactor excelled on hardwood floors, reaching the highest score across all tests with 99.27% sand pickup. Yet, this affordable robot avoided only one out of six items in the avoidance test. In its first trial, the Roomba 205 failed to avoid the sock, the pet toy and all three types of pet waste.
The Dreame X50 Ultra (58.64%) and the Dreame D10S Ultra (62.32%) both rank among the top overall cleaners but avoided only three and four items, respectively, out of the six obstacles.
Units that performed the best at obstacle avoidance have noticeably lower cleaning scores. Across all 24 vacuums, four units tied for the best avoidance score (five out of six avoided): the Yeedi M14 Plus, Eufy E28, SwitchBot S20 Auto-Fill & Drain and Eufy E25. The average pickup for these models was significantly lower than that of top cleaners, such as the Eufy E28 (44.49%) and the SwitchBot S20 Auto-Fill & Drain (25.17%).
The highest cleaning scores are consistently achieved by units that perform poorly in obstacle avoidance, which is crucial for guarding against items such as socks, cords or pet waste.
The units that deliver the highest average cleaning scores often fall short when it comes to navigating a messy real-world environment. What this all means is that purchasing a robot vacuum requires choosing a specialization: You can either get a highly effective cleaner or one that’s better at avoiding objects.
“From a systems perspective, robot vacuums struggle with obstacle avoidance because they’re balancing power, sensor load and path-planning all at once — similar to how HVAC systems can’t optimize airflow and humidity control simultaneously without trade-offs,” said Ahmed Harhara, Ph.D., engineer and founder of HoustonHomeTools.com — a data platform that helps residents understand neighborhood-level environmental and housing risks. “Increasing suction (like in pet mode) draws more power away from sensors or processing, reducing the time or accuracy available for object detection.”
A unique standout: The Yeedi M14 Plus
The only real exception to this was the Yeedi M14 Plus, which managed to avoid five obstacles, while still scoring a 50.03% average pickup, so if you want a robot vacuum with the fewest compromises, this is your best option, with the caveat that it’s not as effective a cleaner as the Mova V50 Ultra Complete. It’ll only cost you $600 at full price, making it a compelling and affordable alternative to the $1,399 Mova V50.
Strong object avoidance robots don’t do as well at cleaning
Four units tied for the best object avoidance score, safely navigating around five out of six obstacles: the Yeedi M14 Plus, Eufy E28, SwitchBot S20 Auto-Fill & Drain and the Eufy E25.
The Eufy E28 only failed to avoid a pet toy, which was also true of the E25, but it was able to avoid pet waste, which is more vital to avoid in our opinion. On the downside, pickup was 44.49% for the E28 and 44.49% for the E25, ranking it lower than our top-performing robots, such as the Mova V50 and Dreame D10S Ultra, which score higher than 60%. However, the E28 may make up for it with a unique selling point by incorporating a portable spot cleaner that can help you clean carpets more effectively.
Even superior navigation doesn’t guarantee complete safety. The SwitchBot S20 Auto-Fill & Drain, despite scoring five out of six in avoidance, experienced a critical failure when 180-degree pet waste wrapped around its wheel. This risk is highlighted by manufacturers themselves. The Noesis Florio’s app even includes a pet mode that enhances pet hair removal but explicitly discloses that it “may reduce the robot’s obstacle avoidance capability.”
“Pet modes deliberately prioritize cleaning over caution,” said Adrian Dunkley, a scientific researcher in AI and applied physics and founder of StarApple AI. “They increase suction, make brushes more aggressive and often adjust movement patterns to maximize hair pickup. Fast airflows can pull in lightweight objects before a sensor can react, and faster or straighter movement reduces the time available for obstacle checks. From an engineering perspective, the software relaxes avoidance thresholds, so the robot cleans more effectively. The trade-off is a design feature, reduced obstacle avoidance for better cleaning.”
Many robot vacuums that claim to have advanced obstacle avoidance fail the hardest
A significant number of units tested demonstrated poor object avoidance, often contradicting their advanced marketing claims. Several units avoided none of the six obstacles (including pet waste, sock, and lamp cord) in the test. Seven units scored 0 out of 6: the 3i S10 Ultra, Narwal Flow, Eureka J15, 3i P10 Ultra, Monsga MS1 Max, Roomba 705 Max and SwitchBot S20 (the model without the auto-fill dock).
The 3i S10 Ultra, which claimed to use “advanced AI algorithms that achieve perfect obstacle avoidance,” actually ran over all six obstacles. The Eureka J15 Pro Ultra also avoided zero obstacles, despite marketing claims for advanced technology. This unit also demonstrated exceptionally weak cleaning performance, recording the worst sand pickup score across all tests on all floor types, achieving just 1.74% on mid-pile carpet. The Noesis Florio says “Pioneering Neural Vision obstacle avoidance,” but avoided only two out of six obstacles.
Running over items led to critical failures, such as the SwitchBot S20 Auto-Fill & Drain experiencing a critical failure when 180-degree pet waste wrapped around its wheel, or the 3i S10 Ultra running over its own base station cord. Oddly, the regular SwitchBot S20 without the auto-fill system failed to avoid any of the obstacles, but had superior pickup at 46.78%.
Ultimately, what this comes down to is that robot vacuums simply can’t do it all.
“Cleaning and obstacle avoidance compete for the same limited resources and space in the vacuum,” said Dunkley. “Strong cleaning requires high suction, fast movement and close contact with the floor. Avoidance requires slower motion, distance and frequent sensor checks. Household objects, such as cords and toys, are small and irregularly shaped, making it difficult for the vacuum to detect them reliably. When a robot prioritizes cleaning, it has less time to recognize and react to objects. When it slows down to avoid them, the cleaning effectiveness drops, so most robots can’t multitask; it’s either power or agility.”
Advice for consumers: How to pick a robot vacuum without too many compromises
Our comprehensive lab testing of 24 robot vacuums reveals a performance trade-off for consumers. We found a consistent pattern that, for the vast majority of robot vacuums, prioritizing raw cleaning often means settling for unreliable object avoidance, which increases the risk of critical failures, such as running over cords or pet waste.
For most consumers, this means prioritizing their needs. Do you want powerful cleaning or better object avoidance? If your priority is deep cleaning (especially on carpets), you’ll want to consider raw pickup scores across each flooring type, in addition to the average. The Mova V50 is the clear standout for mid-pile carpet performance (47.54%) and has the highest average pickup score (65.14%), but you’ll need to baby it a bit more compared with the Yeedi M14 Plus and Eufy E28 by clearing the floor of socks, cords and other small items.
The robot vacuum with the best object avoidance and cleaning performance proved to be the Yeedi M14 Plus, which avoided five obstacles and achieved a 50.03% average pickup, although this is still markedly below our top performers.
For Dunkley, the problem lies on the software end, rather than the hardware end. “Cameras, lidar, infrared sensors and even touch sensors each have strengths and blind spots,” he said. “What remains difficult is combining data from all those sensors in near-real-time to make quick decisions without draining the battery or compromising vacuum performance. Better on-board processing and smarter coordination between sensors and movement will matter more than any single new technology.”
“Most units use either lidar or vision sensors, but real-time clutter recognition requires a fusion of both — plus faster onboard processing,” said Harhara. “For homeowners, the best fix isn’t waiting for tech to catch up — it’s pre-mapping rooms and reducing loose clutter. Just like airflow efficiency depends on clean filters and open vents, smooth navigation depends on predictable surfaces.”
Both Dunkley and Harhara agreed that the key to improving a robot vacuum’s object avoidance and cleaning is to reduce the uncertainty in your home. That means moving cords and oddly shaped objects before the robot starts cleaning, having it clean when the light is good so the vision system is more effective and using deep clean mode selectively in favor of multiple cleaning passes for better coverage.
Correction, Jan. 9: An earlier version of this article misidentified the lab results for one of the robot vacuum models. We have updated it to reflect the correct model name, which is the Mova V50.
Read the full article here
