No audio available for this content.
What role do you think AI will play in signal processing and positioning accuracy?
“In my view, artificial intelligence (AI) is just that, and is only as intelligent as the data it ‘learns’ from. For signal processing and positioning/navigation operations, AI must function in ‘real-world’ physical spaces and dynamic and at times unpredictable environments along with other independent activities. In these cases, past data may not be sufficient to inform current dynamic operations, as the AI-enabled actor must have real-time access to precise positioning and timing information from outside itself to function. As I said in this forum in March, AI won’t be useful in such instances without ‘independent awareness’ of its physical location and timing to synchronize or deconflict its operation with other dynamic activities. In that sense, PNT sources do not depend on AI, but ‘dynamic’ AI must have, and will not work without, reliable PNT.”
— Jules McNeff, Overlook Systems Technologies
“AI continues to grow and permeate all aspects of society. (Unfortunately) for many, it has become more than a passive tool. It has ‘infiltrated’ into areas without yet going through the needed levels of scrutiny to ensure products that support PNT use cases that deliver safety, security, and economic well-being. This is not to say that prudent, thoughtful and proper use of AI cannot help promote ‘better’ PNT services. As with all development in this area, the concepts of ‘trust, but first verify’ must remain a firm, unshakeable pillar of all engineering design. The question for both developers, and most importantly for engineering standards organization, ‘How will the risk of AI products be properly assessed? How will required levels of accuracy, availability, integrity and continuity be ensured?’ What additional testing and test methodologies (on paper and in the real world) will be needed? Will AI be used both on the development side and the certification side? These are gnarly questions that need to be addressed, and doing so properly will take significant time and effort. Somehow I feel we’re in for some ‘black swan events’ down the road if we all don’t agree to hold hands and walk down this path together.”
— Mitch Narins, Strategic Synergies
AI has arrived in the signal processing domain as a hardware designer between the signal specification and the physical implementation (RTL) to reduce the design time.
The lower bound of positioning accuracy depends on the measurement and corrections precision. The tails of the error distributions are driven by the outliers in these data. AI is stepping into the identification of outliers and classification of system performance, even across a fleet of drones experiencing common-mode disruptions from jamming and spoofing.
— Paul McBurney, oneNav
