Neural Propulsion Systems Claims Significant Radar Breakthrough

March 9, 2024

Chris Clonts

Behrooz Rezvani, founder and CEO of Neural Propulsion Systems, cuts to the chase quickly. "We can improve the performance of any radar and help it see clearer, farther and sooner," he said. Using a mathematical framework initially discussed in an MIT research paper 14 years ago, Rezvani says his company can take any manufacturer's radar unit and help it:

• Increase resolution by a factor of 10 for two-dimensional imaging
• Suppress 10 times the number of false positives
• Detect targets at twice the current distance with a lidar-like point-cloud density
• Differentiate notoriously difficult targets, such as pedestrians walking or standing next to parked vehicles

NPS Executive Consultant Lawrence Burns, the former head of GM research and development, has seen plenty of advancements during deep involvement with the development of night vision and adaptive cruise control. But he always knew existing radar systems were not yet the answer for the future needs of hands-free driving and other features.

Rezvani said that the company's algorithm allows for signal parsing at very near the mathematical limits. "At MIT, there was a breakthrough that gets us to the theoretical limits of parsing," he said. That breakthrough recognized the usefulness of frameworks called atomic norms, first discussed in a 2010 paper later updated in 2012.

Ben Recht, now a professor of electrical engineering and computer science at UC Berkeley and one of the original paper's authors, wrote this  as an introduction to the topic: "Many signals and systems that we commonly acquire and analyze can be expressed as linear combinations of a few basic building blocks. For example, radar signals can be decomposed into a sum of elementary propagating waves. Atomic norms provide a framework for estimating these sorts of signals with very few sensors or very fast acquisition times by solving convex optimization problems.

"In the atomic norm project, we are interested in exploring the fundamental limitations of optimization methods in data analysis and how these theoretical techniques mesh with the myriad of complex systems and signals that we encounter in practice."

NPS founder Rezvani followed by saying the company's software could optimize other sensors, including lidar. But radar tech is cheaper – and NPS software can work with fewer chips.

Babek Hassibi, the company's CTO, explained the difference in the software: "The breakthrough here is a totally new way of looking at radar signal processing. The algorithm looks at the received signal and says, 'If I wanted to describe this as a collection of reflections of point targets, where in space are they? What are the distances? What is the azimuth? What are the elevation angles? If there's motion, what are the speeds? It does it all at once.

“Lidar people like to talk about resolution. What many don't talk about is ghosts," he added. "False positives can lead to emergency braking, which can lead to bad things (such as rear-end accidents). How do you relate measurement to what is critical to safety? With radar, the larger the aperture, the narrower the beam you can create to determine whether something is one object, two objects, et cetera.

"What you can't do is miss something. A car, pedestrian or bicyclist," he continued, making the point that one of NPS' triumphs is being able to consistently detect a pedestrian next to a vehicle. "That's one of the instances where someone is trying to change a tire. It's an actual specification of some of the safety requirements." He says the software is six times better than a baseline system at detecting a human and also is adept at identifying two separate pedestrians instead of indicating only one target.

"This is what I think is going to break the logjam of mass deployment. You can get [an angular resolution of] 0.5 degrees or less and it can be affordable. Every asset that we can check says 'yes,' this is the answer."

From a business standpoint, Rezvani said that despite use of a public concept, he isn't worried about being overtaken by another company. "No other company can leapfrog our tech because there is a fundamental limit to how close you can get to those theoretical limits," he said.

The company is working with and has an investment from an OEM it can't disclose and says the U.S. Army just approved a research proposal for a defense application the company is exploring in partnership with Raytheon.

---