Enabling or aspirational? Zoox's grant US12649487B2, "Driving surface cost landscape" (issued June 9, 2026), is firmly enabling, and the classifications say so: B60W 60/001 (autonomous operation), with B60W 2552/05 and B60W 2556/45 marking road-surface and map-data inputs. This is a planning patent — the layer that decides, given everything the car perceives, where on the road it should actually put itself.
The mechanism is a cost landscape, and it is one of the foundational ideas in robot motion planning made concrete for driving. Instead of treating the road as simply drivable-or-not, the method assigns a cost to each location on the drivable surface: high cost near a curb or a detected obstacle, lower cost in the center of the correct lane, intermediate cost where the map or rules make a position less desirable. The planner then searches for the lowest-cost path. Turning the world into a cost surface is how an AV converts a messy scene into a single optimization it can solve.
Why this is the heart of the stack. Perception tells the car what is around it; prediction tells it what those things will do; but planning has to commit to an actual trajectory, and the cost landscape is the representation that makes that commitment tractable. Get the costs right and the car drives smoothly, keeps margin from hazards, and stays where it belongs. Get them wrong and the car hugs curbs or hesitates. The grant claims a specific way of building that landscape from the available inputs.
Novel where the construction is novel. Cost maps and cost grids are deeply established in robotics, so the defensible position is the particular method of constructing the driving-surface cost landscape — which factors, how they are weighted and combined into the surface. The independent claim defines that construction; the dependent claims and the B60W 2552/2556 input classifications fence in the specifics. Reading the abstract as "Zoox owns cost-based planning" would badly overstate the scope.
The assignee is Zoox, Inc. — the Amazon-owned robotaxi developer — and Zoox is, by patent count, one of the most prolific assignees in the autonomous-driving classes. A planning patent from Zoox is part of a dense, deliberately built portfolio; the company treats its motion-planning methods as core defensible IP. For a portfolio analyst, Zoox's concentration in B60W 60/001 is one of the standout assignee clusters in the entire autonomy record.
For the B60W autonomy beat, this grant is a clean example of the enabling-claim category: not a vision of self-driving, but a specific, implementable method for the decision that every AV must make thousands of times a minute. The cost landscape is where perception becomes action, and a granted method for building it is a real position in the autonomy IP race.