A published patent application is a delayed look at where a company spent its research effort, and the volume and consistency of a week's filings can say as much as any single one. In the week of April 28, 2026, Oshkosh Corporation — the maker of refuse trucks, aerial work platforms, military and airport vehicles — had a notably large cluster of applications publish, and a clear seam runs through them: autonomy and machine vision, aimed not at the passenger car but at the working vehicle. The filings point to Oshkosh investing in automating the unglamorous machines that already do industrial work, a different corner of the autonomy map than the robotaxi.

The center of gravity is the refuse truck. US20260116654A1, on an autonomous refuse vehicle with a robotic-arm state control architecture, describes a controller that coordinates the truck's drive motor with its lift system to collect refuse from a container it has detected nearby.

An autonomous refuse vehicle can include a chassis, a body assembly, a drive motor, a lift system, and a controller.— Autonomous refuse vehicle with robotic arm state control architecture, US20260116654A1

That is a robotics filing in plain clothes: a vehicle that drives itself along a route and operates a robotic arm to do a physical task, classified in B65F refuse-vehicle classes alongside B60W 60/00 autonomous-operation control. It is the humanoid-and-robotaxi conversation transplanted onto a garbage truck — arguably a more tractable autonomy problem, because the route is fixed and the task is repetitive.

A cluster aimed at the work vehicle, not the road car

The companion filings fill in the system around that truck. US20260116656A1 describes controller priority for a cart-detection system, arbitrating between an autonomous refuse-collection command and a manual override so that a human signal takes precedence over the automated one — a safety-and-handoff disclosure that only matters once the vehicle is genuinely operating itself. US20260118890A1, on autonomous navigation to an operating position, describes a work vehicle that charges itself and then drives back to the exact position and orientation where its operator left it, ready for the next shift.

The perception and routing pieces round out the cluster. US20260116587A1 describes a ground-condition monitoring and route-generating system that uses camera input — including from a drone — to survey a work site, identify surface conditions and obstacles, and generate a route map for a work machine. And US20260116725A1, on image recognition for lanyard detection, uses a camera to track an operator's position relative to an aerial work platform and move the platform according to where the operator goes. Read together, the week's filings describe self-driving vocational vehicles, the site-mapping that routes them, and the machine vision that keeps their human operators safe.

The breadth across product lines is itself part of the signal. The refuse-vehicle filings (US20260116654A1, US20260116656A1) sit in B65F classes; the aerial-work-platform and telehandler filings (US20260116725A1, and the lift and fork-alignment applications that publish alongside them) sit in B66F lifting classes; the navigation and route-mapping filings (US20260118890A1, US20260116587A1) sit in G05D and G01C autonomy and mapping classes; and a parallel run of charging and electric-drive filings sits in B60L. A company exploring autonomy in only one product would file in one of these neighborhoods. Oshkosh's week spans all of them at once, which reads less like a single experimental program and more like autonomy and machine vision being threaded through the whole vocational portfolio — refuse, access equipment, material handling — in parallel. The same handful of inventor names recurring across the refuse and lift filings suggests a shared internal team carrying common autonomy building blocks from one product family to the next rather than isolated efforts.

The arbitration filing US20260116656A1 deserves a second look in that light, because it is the kind of disclosure a company writes only when it expects real autonomous operation. Deciding that a manual override must take precedence over an autonomous collection command is a safety-governance question that does not arise until a machine is actually driving and acting on its own; it is the sort of detail that surfaces when a program is moving from concept toward fielded behavior, not when it is sketching possibilities.

Where the filings point

The direction here is distinctive, and it is visible in the classification mix. Where the broad autonomous-driving application set that week clustered in passenger-vehicle control classes — B60W 60/001, the B60W 2420 sensor classes, G06V vision classes for road scenes — Oshkosh's filings sit in B65F (refuse vehicles), B66F (lifting and work platforms), and B60L (electric-vehicle propulsion and charging), with autonomy and vision classes layered on top. That is the signature of a company applying autonomy to its existing vocational product lines rather than entering the robotaxi race. The recurrence of the same inventors across the lift and refuse filings reinforces that this is a sustained internal program, not scattered one-off ideas.

For a manufacturer whose business is selling specialized work vehicles to municipalities, contractors, airports and defense customers, disclosed R&D in self-driving refuse trucks and self-navigating work machines points to investment in automating labor-intensive, route-based industrial tasks — the kind of work where a fixed route, a repetitive motion, and a controlled site make autonomy more approachable than open-road driving. The filings also pair autonomy with electrification (on-route battery preconditioning, electric telehandlers and refuse vehicles appear in the same week's cluster), suggesting the two threads are being developed together for the same product families.

The standard caveat is load-bearing. A published application reflects work done well before it surfaces and signals an intention to seek coverage, not a shipped product; many such filings never reach commercialization, and nothing in the record says Oshkosh will field an autonomous refuse fleet, win the contracts to deploy one, or productize any single filing. What the published cluster shows is concrete and bounded: across a substantial set of recently published applications, Oshkosh's disclosed research concentrates on autonomy, navigation, site-mapping and machine vision applied to refuse trucks, lifts and work machines — placing its forward-looking effort on automating vocational and industrial vehicles, a corner of the autonomy field distinct from the passenger robotaxi.