Steer-by-wire's whole proposition is that the driver's wheel and the road wheels stop being mechanically joined. That buys packaging freedom, tunable steering feel and a clean interface for automated driving to command steering directly. It also quietly removes something engineers had been getting for free: a rigid mechanical path that guaranteed a known relationship between what the driver turned and where the rack ended up. Take the linkage out and that relationship becomes a number a computer has to believe in rather than a fact geometry enforces.

An application published on 23 July 2026 and assigned to GM Global Technology Operations LLC is about checking that belief. US20260208784A1, naming Mahdokht Ezati, Seyedeh Asal Nahidi and Amin Habibnejad Korayem, is classified under CPC B62D 5/006, B62D 5/0409 and B62D 5/046 — steering control, all three.

A steer-by-wire system includes a hand wheel actuator coupled to a steering column, a road wheel actuator configured to drive a steering rack with a pinion gear, and a controller in electrical communication with the hand wheel actuator and the road wheel actuator.— System and Method for Identifying Degradation in a Steer-by-Wire System, US20260208784A1

The quantity being watched

The record builds everything on one ratio. It defines a nominal C-factor as the ratio of travel of the steering rack per revolution of the pinion gear — how far the rack slides for one turn of the gear driving it. That is a designed, known quantity for any given rack-and-pinion geometry. The controller determines an actual value for the same ratio during operation and compares the two.

A note on notation, because the published text is inconsistent about it: the record renders the estimated value as Ĉfactor, with a circumflex and no separator, the designed value as Cfactor, and the comparison as C-factor difference with a hyphen. The circumflex is conventional notation for an estimated rather than measured quantity, and the run-together spelling is a flattened subscript. Four spellings of one term in one document is a typesetting artefact, not four different quantities.

What makes this non-trivial is that the actual value is not directly measurable. There is no sensor reading out rack displacement. Claim 8 recites that the actual value is determined from an estimated steering rack displacement and a pinion angle displacement. Claim 9 says that rack displacement estimate comes from parameters regarding the road wheel actuator, and claim 10 enumerates them: a motor current, a motor voltage, a motor inductance, a motor resistance, and a back electromotive force, together with a mass, a stiffness, a damping and a steering arm length.

That list is the actual mechanism. The first five are electrical quantities the motor controller already has; the last four are mechanical properties of the assembly. Feed both sets into a model of how the motor and the rack interact, and you can infer displacement without measuring it. Claim 11 names the technique — the controller utilises an observer to determine the estimated steering rack displacement, taking estimated rack force and motor voltage as control inputs and road wheel actuator motor torque as a measurable control output. An observer is a model that runs alongside the real system and corrects its own estimate against whatever the system does expose. Here the exposed signal is motor torque.

What it does about it

The comparison feeds a graded response rather than a single warning light. Claim 2 determines degradation status by comparing an absolute value of the difference against a predetermined range. Claim 3 applies a compensation C-factor when the difference falls in a first range — the system silently corrects for the drift and keeps operating. Claim 4 raises an alert in a second range, which claim 5 specifies as a service notification. Claim 6 covers a third range, with claim 7 escalating to a service notification or a drivability notification.

Three bands, three postures: compensate quietly, tell the owner to book service, tell the driver the car is not behaving as designed. That progression is the useful part of the disclosure. It treats degradation as continuous rather than binary, which is the honest way to model wear in a mechanism.

Two things the record does not do, and it is worth being exact about both. It never identifies a cause. Nothing in the abstract or claims says why an actual ratio would fall short of nominal — no wear mechanism, no failure mode, no component named. The claims detect a discrepancy and grade it; diagnosis is left outside. And the detection is deliberately one-directional. Every independent claim — claim 1 for the system, claim 13 for a vehicle, claim 19 for the method — computes the difference only when the actual value is less than nominal. A rack travelling further per pinion revolution than designed is not what these claims are looking for.

The application sits in a drop where twenty records carry a GM assignee, and a striking number are about catching faults rather than adding features. A second steer-by-wire filing manipulates the baseline steering ratio from a steering angle gradient. Others cover monitoring software on lower-integrity hardware to raise its safety-integrity classification, detecting cracks in a transparent structure inside a lidar's field of view, and scoring how complex an external environment is for testing automated driving. Remove the mechanical fallbacks and you inherit the job of proving, continuously, that the electronics still describe reality. This filing is one instrument for that — and it is a published application, not a granted patent, so it establishes a claim to a method and nothing about a shipping vehicle.