What role do artificial intelligence (AI) and machine learning (ML) play in analyzing GNSS signals? How might that evolve?
Ellen Hall
“ML is gaining adoption across many GNSS application areas due to its ability to extract data and classify signal information often within complex operational environments. By combining ML with AI, systems are now able to characterize receiver correlator outputs and ranging residuals, and then fuse this with identified environmental features — all potentially increasing GNSS accuracy, integrity and availability. As AI and ML mature, we can expect to see new novel methods to optimize PNT sensor-fusion engines. This will include the combination of GNSS signals with other sensor signals such as inertial and vision.”
— Ellen Hall
Spirent Federal Systems
Bernard Gruber
“AI will come to the battlefield and I would like to think that AI and ML will play a large part in GNSS solutions and specifically protection from adversaries in the future. As AI can ‘anticipate’ threats (i.e., spoofing, jamming, poor coverage) based upon what it sees and knows one should be able to reduce the cycle time to combat that threat (e.g., find/fix/identify and then target, change frequencies, evade). Seeing this data, ML can adapt to morphing threats as well as ‘fuse’ data from all different domains (air, space, sea and land) to provide solutions.”
— Bernard Gruber
Northrop Grumman
Jules McNeff
“I would like to turn the question around and ask ‘How does GNSS contribute to enabling AI and ML to function in physical space?’ Many AI and ML experts don’t think about this aspect of the technologies. Of course, timing is essential to AI and ML operation, but both must be spatially oriented as well if they are to interact effectively with things in the ‘real world.’ The more complex the interactions, the higher the need for precise, continuous PNT information. Depending on the applications, the relationships can become synergistic.”
— Jules McNeff
Overlook Systems Technologies
Greg Turetzky
“AI and ML have a great opportunity to fundamentally change the way GNSS signals are used for positioning. In particular, the new modernized signals with wider bandwidths and higher chipping rates create a fundamentally richer data set than classic range/range rate measurements. By analyzing the channel response and using AI/ML techniques, the entire signal environment of LOS and NLOS signals can all be used to make more accurate measurements. In fact, in deep urban canyons with appropriate training, it is even possible to accurately position using only multipath signals such that more multipath makes the position more accurate, not less.”
— Greg Turetzky
oneNav
