Building the map is itself the camp choice, and this claim builds the map. NVIDIA's grant US12650313B2, "Lane line creation for high definition maps for autonomous vehicles" (issued June 9, 2026; inventors include Mark Damon Wheeler), claims generating lane-line geometry for high-definition maps. HD maps are the centimeter-accurate priors that mapped-autonomy systems localize against — and a patent on creating them is, by definition, IP for the mapped camp, not the map-free vision-only one.
The CPC stack draws the line cleanly. G01C 21/3638 and G01C 21/3635 cover producing and updating map data for navigation; G01C 21/3867 covers map-matching; B60W 40/04 covers estimating road conditions; and G06V 20/582 and 20/588 cover recognizing lanes and road markings from imagery. That combination describes a pipeline that turns sensor observations into durable map features — the map-authoring side of autonomy, distinct from the live-perception side.
Why this matters as a camp-lines claim: the vision-only-versus-mapped debate is partly a debate about whether you carry a pre-built HD map at all. Tesla's stated philosophy minimizes reliance on them; Waymo-style stacks lean on them heavily. A claim on creating lane lines for HD maps is a vote — and an asset — for the map-dependent architecture. A vision-only system that deliberately drives without an HD map doesn't practice a method whose limitations recite creating HD-map lane geometry.
For claim scope, the lane-line specificity is the boundary. The grant isn't a monopoly on HD maps generally; it's a method for generating one feature class — lane lines — for them. The metes and bounds live in how the claim recites deriving that geometry. A competitor building HD maps by a materially different lane-extraction method may avoid the fence, which is why the recited derivation, not the concept of HD maps, is what an analyst checks.
Caveats, plainly. Map-creation methods are improvement-dense and continuation-prone — expect a family around a claim like this, and read the whole family before calling the scope. The examined B2 grant cleared the bar, but classification breadth isn't claim breadth. And a method to create map features is not evidence anyone's fleet uses it; it's evidence NVIDIA owns the technique.
For the control beat, sort this with the mapped-camp IP. NVIDIA is arming HD-map autonomy with a lane-authoring claim, the same way its Doppler-adjacent and LiDAR perception grants arm the sensor-heavy side. Pull it on PatentBear, confirm the HD-map lane-line limitation in claim 1, and read it as what it is: a claim that only a map-carrying architecture practices.