A granted patent is not a research note; it is enforceable coverage, a claim a company can assert. So when a single U.S. grant cycle issues a dozen patents to one automotive supplier on the same day, the question for a business reader is concrete: what did the company just lock in, and where in the stack does that coverage sit? For Robert Bosch GmbH, whose patents issued under the April 7, 2026 grant date, the answer is legible. Bosch does not build a self-driving car; it sells the radar, the cameras, the localization software and the electronic control units that a carmaker integrates into one. The autonomy-relevant grants this cycle map onto exactly that layer — the sensing-and-software box, not the vehicle around it.

The clearest single filing is US12596367B2, a method for the semi-automated guidance of a motor vehicle that leans on roadside infrastructure. The grant describes generating control signals for lateral and longitudinal guidance without infrastructure support, checking whether the infrastructure-supported function is working, and terminating the infrastructure-assisted mode depending on that check — a fallback-on-loss design built into the claim. Its classification spread (G05D 1/0055, G05D 1/0088, G05D 1/0276) places it in the autonomous-control corner. It is coverage over the handoff between a smart-road system and the car's own controller, the seam where infrastructure-assisted driving either degrades gracefully or fails.

During the infrastructure-supported, at least semi-automated guidance of the motor vehicle, control signals are generated for the at least semi-automated control of a lateral guidance and/or longitudinal guidance of the motor vehicle without infrastructure support.— Method for the semi-automated guidance of a motor vehicle, US12596367B2

The map and the localization layer

Two of the week's grants sit on the map itself. US12596015B2 covers creating a map representation of a road network by receiving environment-sensor data from a fleet of vehicles, running a SLAM method on that data on an external computing unit, then fusing differential-GPS pose information into the map — coverage over the crowd-sourced, fleet-built HD map approach that turns ordinary cars into mapping probes. US12596013B2 pairs with it, covering both the creation of a digital map representing a region along a traffic route and the operation of an automated vehicle along that route, classified in B60W 60/001 (the autonomous-driving conjoint-control class) alongside G06V 20/588 lane-perception classes. Read together, the two grants give Bosch issued coverage over the build-and-use loop of an HD map: collect from the fleet, build centrally, drive on the result.

Localization depends on signals that can be jammed or spoofed, and one grant addresses that directly. US12596198B2, a method for providing GNSS-relevant route information, covers receiving raw GNSS satellite data, forming an "interference indicator" describing an interfering influence on GNSS reception for a route section, and storing that indicator against geographic information. It is coverage over mapping where satellite positioning becomes unreliable — a quiet but load-bearing piece for any system that hands control back to the driver when localization confidence drops.

Sensing, perception robustness and the network underneath

The radar grants are where the supplier identity is most visible. US12596173B2, a radar sensor device and method for self-testing, covers a test-signal path running between a coupling point on the transmit path and one on the receive path, positioned relative to the antennas so the sensor can verify its own integrity (classes G01S 7/032, G01S 7/4017, G01S 13/931). For a component a carmaker bolts on and must trust for the life of the vehicle, built-in self-test is the kind of claim that matters to the buyer. On the perception side, US12597246B2 covers a method for generating adversarial patches for an image by segmenting it, selecting target regions under an attacking criterion, and generating the patch set — a filing in the robustness-testing corner of automotive perception (G06V 10/82, G06V 10/776), where the concern is how a vision model fails when an image is deliberately perturbed.

Underneath all of it is the in-vehicle network, and Bosch issued coverage there too. US12598195B2 covers a method for intercepting an intrusion into a bus system: an intrusion-detection system flags a frame from another node as an intrusion, then sends data on the receive line between transceiver and controller to manipulate the arriving signal — a claim over actively neutralizing a malicious CAN-bus frame, not merely detecting it. Closer to the mechanical edge, US12594921B2 covers an electromechanical brake-pressure generator with an anti-twist protection, part of the brake-by-wire actuation that automated driving ultimately depends on.

The four classes Bosch is fencing across in this cycle — perception (the adversarial-patch and camera-map filings), localization (the SLAM map and GNSS-interference filings), the radar sensor itself, and the network-and-actuation layer (CAN-bus intrusion and brake-pressure generation) — are precisely the pieces that ship as separate part numbers in a supplier's catalog. A carmaker buys a radar here, a localization stack there, an ECU somewhere else. Issuing same-cycle coverage spread across those distinct components, rather than concentrated in one, mirrors the way the business itself is organized: as a set of integratable building blocks. The grant record reads less like one product and more like the catalog, each entry now backed by an issued claim.

Set against Bosch's position as one of the volume sources of automotive-autonomy grants in this week's cycle, the shape of the block is the story. The coverage Bosch issued runs through guidance handoff, the HD map, GNSS interference, radar self-test, perception robustness and the network security beneath them — the pieces of an automated-driving platform that a supplier delivers as components and software, and that a carmaker integrates rather than invents. None of these grants describes a finished autonomous vehicle; collectively they describe enforceable coverage over the sensing-and-software layer that sits inside the ADAS box. For a reader tracking where a Tier-1 supplier is fencing, this week's claims landed on the platform layer between the road, the sensors and the controller — and each is now issued and assertable rather than merely pending.