A conventional automotive lidar fires one laser wavelength and times the reflections to build a depth map. It tells you where surfaces are. What it does not tell you, easily, is what those surfaces are — a wet road and a dry one, a dark car and shadow, can look similar in single-wavelength returns. A 2023 GM grant takes a different approach: use multiple wavelengths to extract more than just distance.
The record: on July 11, 2023, GM Global Technology Operations LLC was granted US11698445B2, “Automotive Lidar with multi-spectral depth imaging and discrete scanning mechanism.” The CPC classes are lidar-specific — G01S 7/4817 (optical scanning), 7/499 (signal processing), and 17/931 (lidar for road vehicles). Two ideas are in the title: multi-spectral imaging and a discrete scanning mechanism.
Here is what multi-spectral buys you. Different materials reflect different wavelengths by different amounts — a surface might look bright at one wavelength and dim at another. By measuring returns at several wavelengths, the sensor gathers a kind of spectral signature alongside the depth. In principle that helps distinguish materials and surfaces a single-wavelength lidar would lump together, enriching the scene the perception system has to interpret.
The “discrete scanning mechanism” is the other half. Lidar has to sweep its beam across the scene; how it scans affects resolution, frame rate, and reliability. A discrete (versus continuous) scanning approach is a design choice about how the beam steps through the field of view. Pairing a richer signal (multi-spectral) with a particular scanning method is GM tuning both what the sensor measures and how it covers the scene.
The trade-off is honest to state. Multi-spectral imaging adds capability but also complexity and cost — more wavelengths means more emitters, detectors, and processing. A single-wavelength lidar is simpler and cheaper. GM patenting the multi-spectral approach is a bet that the extra discrimination is worth the extra hardware, at least for some applications. Whether it ships, and where, the grant does not say.
The skeptic's caveat, as always with sensor patents: a granted method for multi-spectral depth imaging is a claim on a technique, not a deployed, validated capability. Richer sensor data still has to be fused and acted on correctly to matter. But it is a real and interesting direction — getting lidar to report not just where things are but something about what they are — and a 2023 GM grant marks the incumbent staking a position on it.