An issued patent is enforceable coverage, which makes a week's grant list a usable map of where a company has just locked something in. In grants issued the week of April 21, 2026, Toyota Motor (TM) was the volume-leading assignee among autonomous-vehicle filings in the records reviewed, with 66 grants under its various entities. Read by subject matter rather than by count, those grants fall into two distinct buckets, and the split is the story: roughly one set fences off camera-based road perception for driver assistance, and another fences off the control of vehicles that move without a driver — not on public roads, but inside Toyota's own plants.

The road-perception bucket

The first cluster concerns what a forward camera sees and how the system reacts. US12608957B2 covers detecting that a preceding vehicle is decelerating by counting the painted dashed lane lines between the lead car and a fixed reference point in a monocular camera image — a low-cost way to infer a slowdown from a single camera rather than radar. US12609033B2 is a driving-assistance claim about alerting occupants when a solid object appears in an adjacent lane within a defined area ahead and closes to within a threshold distance. Both sit in CPC G06V 20/58 and 20/588, the road-scene and lane-marking recognition classes, and both are framed around assisting a human driver rather than replacing one.

Two more grants in this bucket are about the machine-learning plumbing behind perception rather than a specific alert. US12608953B2 claims using the discriminator from a generative adversarial network to flag anomalies in roadway sensor data, and US12608951B2 covers generating ground-truth training data for one sensor's recognizer by projecting another sensor's recognition result into its coordinate frame. The latter abstract states its mechanism plainly.

The first recognition result is projected onto the second surrounding environment data by transformation from a coordinate system of the first sensor to a coordinate system of the second sensor.— Method and system for generating ground truth data for machine learning of recognizer, US12608951B2

Claims on data labeling and anomaly detection are notable in a coverage sense because they sit upstream of any particular product feature: they cover the process of building and validating the perception models, not just one detector. That positions the grants across a wider slice of the development pipeline than a single ADAS function would.

The unmanned-movement bucket

The second cluster is easy to misread as robotaxi work; the records suggest it is about manufacturing. US12608021B2 covers controlling a movable body that travels by unmanned driving to lengthen or shorten its travel time according to how the traveling environment affects the movable body's quality — language tied to vehicles being driven autonomously through a factory before delivery. US12608011B2 is sharper still: it claims disabling remote control of a moving object once it detects a disablement signal at a predetermined place along its route, explicitly during the course of manufacture in a factory. US12608015B2 rounds it out with a control device that estimates a movable body's position and orientation from a selected sub-range of three-dimensional point-cloud data. All three carry G05D control classes and factory-context CPC tags (G05D 2107/70), separating them cleanly from the road-facing perception grants.

The factory-movement claims are worth dwelling on because they are the part of the map a general reader is least likely to expect from an autonomous-vehicle grant list. A finished car has to get from the end of the assembly line to a holding lot and onto a transport truck, and doing that with vehicles that drive themselves — under remote supervision, on private property — is a logistics problem distinct from public-road autonomy. The disablement claim in US12608011B2 makes the setting unambiguous: it covers turning off remote control at a predetermined place along the moving object's route during the course of manufacture in a factory, which is a hand-off-to-human-control event at the boundary of the plant. Read alongside the travel-time-versus-quality claim of US12608021B2 and the point-cloud localization of US12608015B2, these three grants describe coverage over moving inventory autonomously inside a controlled environment — a use case that shares perception and control techniques with road autonomy but answers to entirely different constraints. That an automaker is fencing off this in-plant layer at all is the kind of detail a coverage map exists to surface.

What the two buckets share is that they are both about removing or assisting the driver, but at opposite ends of the vehicle's life: the road-perception claims govern the car a customer operates, while the unmanned-movement claims govern the car as inventory the plant moves on its own. A company issuing in both areas in the same week is building coverage that spans assisted driving and its own manufacturing logistics, rather than concentrating on a single robotaxi-style application. That breadth is itself the map — the grants describe a manufacturer extending automation claims into the factory as well as onto the road.

Two framing caveats keep this grounded. First, grant timing reflects prosecution that began years earlier, so this week's issue list describes coverage Toyota sought well before now, not a fresh strategic pivot. Second, a granted claim defines what an assignee can assert; it does not by itself indicate a shipping feature or a deployed factory system, and the records here are claims, not products. What the issued patents do establish concretely is the shape of the coverage: in a single week, the volume-leading autonomous-vehicle assignee added enforceable claims across both camera-based road perception and unmanned in-plant vehicle movement, with the company on the assignee line of each. The competitive read a general reader can take from that is narrow and factual — Toyota's freshly issued automation coverage this week was split across the consumer car and the factory floor, and other assignees in the same window (NVIDIA and Hyundai among them, by grant count) were filing in adjacent road-perception classes rather than in the factory-movement ones.