SLAM stands for simultaneous localization and mapping. It is the foundational robotics problem of building a map of an environment the robot has never seen before while, at the same moment, estimating where the robot is inside that map. The two tasks are circular — you need a map to know where you are, and you need to know where you are to build a map — and SLAM is the family of algorithms that solves them jointly. Every autonomous floor cleaner, warehouse robot, and self-driving prototype that operates without a pre-loaded high-definition map is doing some form of SLAM. In the patent record, SLAM-type inventions sit in the navigation-and-control families, most often under CPC G05D, whose subclass title is "SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES," with G05D 1/00 directed to "Control of position, course, altitude or attitude of land, water, air or space vehicles."

The first thing to know about reading SLAM patents is that the claims almost never say "SLAM." Claim language is built from concrete, infringeable steps, not from the name of the academic problem. What you look for instead is the signature trio: receiving sensor data, generating or updating map data from it, and localizing the device relative to that map. US9612123B1, "Adaptive mapping to navigate autonomous vehicles responsive to physical environment changes," a granted patent assigned to Zoox Inc and directed at exactly this problem, recites the pattern plainly in its claim 1.

aligning, by the computing system and for the particular driverless vehicle, the aligned sensor data relative to positioning map data pertaining to a global coordinate system to localize the aligned sensor data to a location of the particular driverless vehicle; generating, by the computing system, data sets of three-dimensional map data based on the aligned sensor data and the positioning map data— US9612123B1, claim 1, source

Read those two steps together and the SLAM structure is visible without the acronym. The claim localizes sensor data to a position in a global coordinate system, and it generates three-dimensional map data from that same aligned sensor data — position and map produced from one pipeline. The claim goes further into the fleet dimension: it receives sensor data from "multiple driverless vehicles in a fleet," detects a change between the newly generated map and a stored map representing "a state change in the environment," updates the stored map, and transmits the updated map back to the fleet. That is collaborative, change-aware mapping — the map is not static, and the position estimate must hold even as the world changes. The claim is directed to keeping a shared map current, which is the operational hard part of SLAM at fleet scale.

SLAM on smaller robots: the virtual-wall pattern

SLAM is not only a self-driving problem. Consumer and warehouse robots run it on cheaper sensors, and the patent record reflects that. US10967512B2, "Moving robot and controlling method," a granted patent assigned to LG Electronics Inc, claims a system in which "an autonomous cleaner which travels in a cleaning area" includes "a controller that generates a map of the cleaning area, and manages a movement of the autonomous cleaner based on the map." The claim then adds a layer on top of the generated map: a terminal that "sets at least one virtual wall in the map" and assigns attributes to locations, so the robot "divides a travelable area based on the virtual wall" and "performs a specified operation according to the attribute assigned to the virtual wall when reaching the location." The mapping is the SLAM substrate; the virtual-wall layer is the claimed improvement built on top of it. This is typical — many robotics patents take the existence of a self-generated map as the foundation and claim something narrower that operates on that map.

The distinction matters for anyone assessing a SLAM portfolio. A claim directed to the core localization-and-mapping pipeline (sensor alignment, map generation, change detection) fences off the substrate itself. A claim directed to an operation performed on an already-generated map (virtual walls, zone attributes, cleaning behaviors) fences off a feature, not the foundation. Both are SLAM-adjacent; only the former reaches the mapping engine. Reading the independent claim is how you tell which kind you are holding.

Why SLAM claims age the way they do

The fleet dimension in US9612123B1 also points at where modern SLAM claims concentrate their novelty. Single-robot SLAM — one device building one map of one space — is the textbook version, and its core algorithms are decades old. The harder, more recently contested problem is keeping a shared map consistent across many vehicles operating in the same environment, detecting when the world has changed, and propagating the update so the whole fleet benefits. The claim's recitation of receiving data from "multiple driverless vehicles in a fleet," "detecting a change... representative of a state change in the environment," and "transmitting the updated data set... to the multiple driverless vehicles" is directed at exactly that collaborative-and-change-aware layer. For an IP reader, this is the signal that a SLAM claim is reaching for something beyond the public-domain core: the novelty is rarely in "build a map and localize," which is old, but in the coordination, change-detection, or update protocol layered on top.

SLAM has been an active research field since the late 1980s, which means the foundational ideas — Kalman-filter and particle-filter localization, loop closure, graph-based map optimization — are decades old and largely in the public domain. Patents filed today therefore tend to claim specific implementations: a particular sensor combination, a fleet-coordination mechanism, a change-detection trigger, a map-update protocol, or an application-specific layer like the virtual-wall scheme above. That is why SLAM claims tend to be narrow and implementation-specific rather than broad and foundational, and why the navigation-and-control class G05D is dense with incremental filings. For an IP reader, the takeaway is to discount the title and the marketing: a patent titled for "autonomous navigation" or "mapping" is doing SLAM, but what it actually fences off lives in the independent claim's specific recited steps — receive, generate or update, localize — and in whatever narrowing limitation the applicant added to clear the decades of prior art that came before it.