January 8, 2026 - Cleaning Up Radar Noise

This is the Telemetry Transportation Daily for January 8, 2026, and I'm Sam Abuelsamid, Vice President of Market Research for Telemetry.  

Automakers have been using radar sensors on vehicles for more than two decades as a core part of their driver assistance systems, particularly for adaptive cruise control. Radar has some distinct advantages over cameras, most notably that it can detect in total darkness as well as through atmospheric obstructions like fog, rain, or snow. Radar can also accurately measure the distance to objects and their relative speed. However, the low-cost radar used on most vehicles is low in resolution and is subject to a lot of scattered returns that can lead to false detections. As a result, the signal processing does a lot of signal processing to reduce the noise and ensure that only targets with a high probability of being real are passed on to control systems. Thus, a pedestrian standing next to a truck will often not be detected because of the noise and few real returns compared to the truck itself. 

A number of companies are now doing more sophisticated radar signal processing to extract more relevant data to improve vehicle safety. Aptive has applied machine learning to its latest generation of radar sensors to get more details. Oculii software from Ambarella generates unique waveforms to get higher resolution returns. Startup Atomathic is using clever AI algorithms to do more intense analysis of the returns, trying to recognize the returns multiple times for each frame. With each pass, it gets slightly different results, but those that keep repeating have a much higher probability of being real targets. As a result, with the Atomathic approach, it can indeed distinguish a pedestrian standing just a foot or two from a truck. 

It takes about five to six times the normal amount of radar processing to achieve this, but with only 16 channels from a typical ADAS radar, it's still at least a couple of orders of magnitude less processing than it takes to analyze a single camera frame. Dramatically improved radar performance is possible with no new hardware in most cases. 

All of these new techniques for radar processing can be utilized to significantly improve safety, even with the current standard entry-level radar systems. 

Thanks for listening.