A SLAM patent's classification tells you which robot it's protecting before you read a word of the claim. iRobot's grant US10705535B2 ("Systems and methods for performing simultaneous localization and mapping using machine vision systems," issued July 7, 2020; inventors Mario Munich, Nikolai Romanov, Dhiraj Goel, Philip Fong) pairs vision-based control codes with B25J 5/007 — the bucket for mobile robot bases. This is the navigation brain of a floor robot, fenced as such.
The mechanism is camera-driven SLAM: the robot builds a map of a space and tracks its own pose within that map using machine-vision features rather than LiDAR or beacons. G01C 21/206 (constructing maps for navigation) plus G05D 1/0246 (control by visual means) is the textbook signature of visual SLAM, and the combination with the mobile-base codes makes clear the host is a self-propelled consumer robot.
For the control beat, the value here is that iRobot is one of the few companies with a real installed base running SLAM in millions of homes, which makes its mapping IP unusually well-grounded. A claim from iRobot on visual SLAM isn't speculative — it fences a technique that ships. That commercialization is exactly the corroborating context the patent desk looks for: the claim describes something that demonstrably works at scale.
From a portfolio angle, iRobot's navigation patents are the moat under a commodity-looking product. Anyone building a camera-mapping floor robot has to navigate around this family, and the inventor list — long-tenured iRobot mapping engineers — signals a sustained, deliberate portfolio rather than a one-off. That depth is what makes consumer-robotics IP defensible even when the hardware looks cheap.
Caveats. Visual SLAM is a dense prior-art field stretching back to academic monocular-SLAM work; the granted claim almost certainly turns on a specific feature-extraction or map-update step, not on "do SLAM with a camera." The classification points at the domain; claim 1's recited operations are the fence.
For the file: a well-commercialized, correctly-classified consumer-robot visual-SLAM grant. Pull US10705535B2 on PatentBear, read claim 1 for the specific mapping or relocalization step it recites, and treat it as a marker of how deep iRobot's navigation moat actually runs.