A "Geiger-mode" lidar sensor sends out pulses at a high repetition rate (200 kHz), forming an image on the percent of pulses that return. The technology has been used by vehicle manufacturers to support collision avoidance, adaptive cruise control, and other Advanced Driver Assistance System (ADAS) applications. But how will factors like snow or another vehicle’s lidar impact a sensor's reading?
In a presentation titled Sensor Fusion and ADAS, a reader asked a supplier of Geiger-mode lidar technology:
For ADAS-related imaging, what protections are available to filter out weather or crosstalk with other vehicles?
Jay Liebowitz, Chief Sales & Marketing Officer, Princeton Lightwave (Cranbury, NJ): Weather is a very interesting question. The fundamental processing technique in Geiger-mode imaging is based on what is called coincidence processing. In this statistical method, you're looking for a pattern in returns so that you can have a confidence level in a measurement. The more often that something occurs in multiple point clouds (since we are doing a very high-repetition sampling of the scene in front the car), the more confidence you have. Weather will provide a random pattern compared to that, so you can actually filter out weather very easily. That’s why [our technology has] been able to look out into the snow.
Regarding crosstalk, this is a question that comes up a lot, because you have lasers on cars that are broadcasting at each other. So how does my lidar know that I'm seeing a scene as opposed to a laser from another car [that makes an object seem] closer than it is, because that laser signal is coming in to my detector. In Geiger mode, you are sampling at a very high rate. In coincidence processing, if you are looking for a statistical coincidence occurring at a high frequency, a very high number of times in an oversampling approach, you can filter out what [data has] a lower level of confidence. That way, Geiger mode lends itself very well to avoiding crosstalk between cars.
What do you think? Share your comments and questions below.