How CNET tests robot vacuums

Our method of evaluating robot vacuum cleaners is straightforward, yet intentionally grueling. Every robot model performs differently, and the performance of each vacuum usually changes depending on the type of surface it's travelling over. Sometimes there are big surprises. That's why we put in the hours and hours of lab time, so we can give buying advice with confidence.

We run tests to evaluate robot vacuums for cleaning power and navigation efficiency. The first trial is to figure out how well a robot covers the floor while it cleans. We built an industry-standard testing room as specified by the International Electrotechnical Commission, just for this purpose. The IEC is an international standards body responsible for managing robot vacuum testing procedures, among other things, for vacuum manufacturers.

Inside this room are objects designed to simulate typical obstacles a robot vac encounters for navigation as it cleans. These obstacles include wall edges, table and chair legs, couches, other furniture and so on, plus bare tile and hardwood floors, as well as carpet.

We mount LED lights on the top of each vacuum cleaner. The dimensions of the lights correspond to the measured nozzle width of each particular robot vacuum that we test.

As the robots move through the room while cleaning, a camera overhead captures a long-exposure image of the entire room in low light. That photo will then have a light trail, created by the LEDs, that shows the exact areas where the robot traveled (and its nozzle position) during its running time. We can also see areas of the floor the vacuum may have missed or gotten stuck.

You can see the navigation results of all the robot vacuums in our test group in the gallery below.

Some robot vacuums have a better sense of direction than others

Recently a few robot vacuum makers have claimed that their machines are smart enough to recognize solid pet waste. To test whether robot vacuums can do this, we place three pieces of prank dog poop inside a walled test pen. Each piece is made from three different materials (hard plastic, soft rubber, silicone). We mark the location of each fake waste pile with painting tape.

We then put each robot inside and record the action from above. If a robot does not come into contact with any of our test turds we consider that a successful run.

The second type of test reveals exactly how much physical debris a vacuum is able to pick up off of the floor. To mimic dirt of small particle size, we use a mixture of play sand and landscaping sand. For bigger particle soil, we use grains of uncooked black rice. Robots then run in straight line mode across three types of flooring (low-pile carpet, medium-pile carpet and hardwood bare floors).

We control for the specific nozzle width of each vacuum, too. We constructed an adjustable tool to soil our test floors. It lets us lay down a strip of precise area of soil to match the nozzle dimensions for every robot. The mass of soil isn't chosen at random either. We measure a proportional amount that's related to the flooring material, type of debris, plus each vacuum's nozzle width.

We conduct three cleaning runs (at minimum) on each floor type. We also perform cleaning tests with sand and rice separately. That comes to at least 18 tests per robot vac. We weigh the robot's dust bin both before and after each run. From there we can calculate the percentage of debris pickup for every cleaning run and the average amount of soil a machine manages to remove. Additionally we run anecdotal (visual) pet hair tests for each robot, on all three floor types.