The sensor modality is in the claim, and it is a meaningful one. Application US20250164644A1, "Control of autonomous vehicle based on environmental object classification determined using phase coherent LIDAR data" (published May 22, 2025), with Aurora founders among the inventors, hinges on phase-coherent — that is, FMCW — lidar. Unlike conventional time-of-flight lidar, FMCW measures each point's radial velocity directly via the Doppler shift, so every return carries motion information, not just position. The CPC anchors it in G01S 17/931 (automotive lidar) and G06V 20/58 (object detection).
That instantaneous velocity is the enabling hook. The claim is about classifying objects using phase-coherent data, and per-point velocity is genuinely useful for classification — a moving pedestrian, a stationary sign, and an oncoming vehicle separate more cleanly when each point reports its speed. This is not aspirational sensor marketing; it is a concrete capability that FMCW lidar uniquely provides, and Aurora has long bet its stack on FMCW lidar specifically. The claim ties that sensor advantage to a classification-and-control method.
The skeptic's discipline: this is an A1 application, so the issued scope will narrow, and FMCW-lidar perception has growing prior art as the sensor type matures. The defensible novelty is the specific classification method that exploits phase-coherent data, and the dependents that pin down how velocity-per-point feeds classification are the moat. The broad independent recitation of 'classification using phase-coherent lidar' will face prior-art pressure.
Strategically, the inventor list and the FMCW focus make this a coherent extension of Aurora's long-standing lidar thesis — the company acquired FMCW lidar capability years earlier and files steadily to protect the perception methods that depend on it. Dated mid-2025, it reflects FMCW lidar moving from differentiator to patented method. The verdict: enabling, because per-point velocity is a real sensor-distinct capability — label it an application, and read the classification-method dependents for the actual scope.