On June 23, 2026 the U.S. Patent and Trademark Office issued a grant to Toyota Jidosha Kabushiki Kaisha that is not, in the end, a headlight patent in the ordinary sense. It is a perception patent that uses the headlight as its actuator. Titled "Vehicle control method, vehicle control system, and non-transitory storage medium" and carried at US12664794B2, it is directed to a method that decides where a traffic light might be and then pushes extra light onto that location so the vehicle's camera stands a better chance of confirming it. The label first, because precision is the point: this is an issued, granted patent — an enforceable position as of the grant date, not a pending application. The question for an IP read is what the granted independent claim requires, and where it sits in the classification landscape.

Start with the term the claim is built around: the signal candidate position. The method begins by recognizing, with a vehicle-mounted sensor, a position where a traffic light is possibly present — a deliberately uncertain object, not a confirmed traffic light but a candidate location. The vehicle carries a light whose distribution state can change: a beam that can be aimed and modulated rather than a fixed pattern. On recognizing the candidate, the claim runs a first light control process: at least one of two things — raising the intensity emitted toward the candidate above its pre-recognition level, or making it higher than the intensity emitted elsewhere. In plain terms, the headlight leans toward the spot where a signal might be. That alone is a recognizable adaptive-lighting idea; what makes the granted claim more than that is the second half.

A vehicle control method comprising: recognizing a signal candidate position around a vehicle by using a sensor mounted on the vehicle, the signal candidate position being a position a traffic light is possibly present, the vehicle including a light configured in such a manner that a light distribution state of the light changes; performing a first light control process of controlling the light in response to recognition of the signal candidate position, the first light control process including at least one of: increasing an intensity of light emitted to the signal candidate position to a value higher than before the recognition of the signal candidate position; or making the intensity of the light emitted to the signal candidate position higher than an intensity of light emitted to an area other than the signal candidate position; acquiring an image of surroundings of the vehicle by using a camera mounted on the vehicle; performing, after the first light control process, a screening process of narrowing down the signal candidate position based on the image; and performing a second light control process of controlling the light after the screening process, wherein an excluded position is a position excluded from the signal candidate position by the screening process; and wherein the second light control process includes reducing an intensity of light emitted to the excluded position to a value lower than before the screening process.— Vehicle control method, vehicle control system, and non-transitory storage medium, US12664794B2

The independent claim is a closed loop, not just a brighter beam

Read as scope, the load-bearing limitations are in the back half. After the first process brightens the candidate spots, the method acquires a camera image, runs a screening process that narrows the candidate positions based on that image, and then runs a second light control process that reduces the intensity aimed at any position the screening excluded, below its pre-screening level. That sequence distinguishes claim 1 from a generic "shine more light at traffic lights" idea: the beam is aimed, then corrected by what the camera reports. As an autonomy-IP matter, this is an enabling-mechanism claim rather than an aspirational one — the limitation that does the work is the screening-and-reduction step, because it binds the actuator to the perception result in a recited order.

The dependent claims fill in how the candidate is recognized, and they read like a catalog of the ways a vision stack guesses at a signal. Claim 9 recites tentatively recognizing a traffic light in the image; recognizing a light source of a signal-indication color; recognizing the red light of a preceding vehicle and setting the candidate above it; or recognizing an intersection from the image or map data. Claim 10 adds a timing wrinkle for two lights at one intersection — illuminating the nearer light when estimated time-to-reach is at or above a threshold, then switching to the farther one once it drops below. Claim 3 narrows to illuminating only the candidate without lighting the surrounding area; claim 5 dims everything that is not a candidate. These dependents refine the candidate-selection and beam-shaping logic; the independent claim 1, and its mirror system claim 11 and storage-medium claim 12, are where the perception-coupled scope lives.

Where it lands in the CPC landscape

The classification tells you which fight the patent is in. The record's lead classes are G06V 20/584 — recognition of objects in a scene specific to traffic lights — and B60Q 1/143 — control of the light distribution of vehicle headlights, the adaptive-beam subclass. Those two CPC homes are the patent's thesis in shorthand: G06V is the camera reading the world for signals; B60Q 1/14x is the headlight changing its distribution in response. A patent that lives in both is, by classification, an actuator-coupled-to-perception filing. Most adaptive-headlight art sits in B60Q alone (glare avoidance, high-beam assist, bending light into curves); most traffic-light-recognition art sits in G06V alone (classifiers, state detection). This grant claims the bridge — using the lighting system as an instrument of the recognition system — and the dual placement is consistent with that bridge being the claimed contribution rather than incidental. It situates the patent at the perception–lighting seam, though it does not, on its own, speak to how the scope compares to prior art. What is factual from the granted text: the enforceable independent claim requires the full loop, and an implementation omitting the reduction-at-excluded-positions step would fall outside claims 1, 11, and 12 as written.

Against Toyota's same-day grant cluster, the hero is the outlier

Against Toyota's other grants in the June 23 drop, this patent is the one pointed at perception while the rest point at the powertrain and the pack. The nearest neighbor in spirit is US12663816B2, "Group control system and group control method," directed to controlling a group of autonomously traveling mobile objects and learning a deep-reinforcement-learning model to correct the deviation between a position sensor's reading and an estimated position — also a perception-and-control filing, but about group positioning rather than signal-aware lighting. From there the cluster turns hardware. US12665356B2, "Charging inlet," is directed to an EV charging inlet with a heater built into the cover member that contacts the charging terminal when closed — a cold-climate port detail. US12665468B2, "Drive device," is directed to an electric-motor drive unit with a counter gear, a hypoid gear, and two separate lubricating-oil volumes. And two records sit in cell and pack construction: US12665271B2, "Secondary battery," is directed to a cylindrical exterior part with cover terminals at both opposing openings integrated with resins, and US12665250B2, "Power storage," is directed to a cell stack held between a pair of recessed casing walls whose central portions press inward against the stack. The throughline is breadth — same-day grants spanning autonomy positioning, charging hardware, drive-unit lubrication, and battery construction — and the headlight patent is the one record where the claimed contribution is a vision-guided control loop. On the face of the granted record, what it covers is an adaptive beam slaved to a two-stage traffic-light screening process; the dual G06V/B60Q classification frames that scope but does not decide it.