We presented our Lake Michigan data collection paper in Portland, Oregon in July at the AP-S/URSI annual meeting. The paper presentation was well-received, including plenty of good questions from the audience. The conference attendance afforded the opportunity to meet and catch up with researchers interested in the same atmospheric ducting phenomenology. In particular, researchers from University of Michigan and Ohio State University gave great feedback and could become collaborators in the future.

A key takeaway from the session and similar sessions is that machine learning is in wide use in propagation modeling. Neural networks require a sizable data set, but can compute outputs such as gain, impedance match, etc., rather quickly. They require less expertise in antenna design and other aspects of the problem to optimize the outcomes. Traditional EM solvers are very computationally complicated, requiring an antenna model, environmental model, and other facets that need to be carefully considered and optimized. The modeling is expertise-driven and the computations are time-consuming because of their complexity. Neural networks provide a reprieve to those burdens.

Another takeaway is the new availability of tools, through MathWorks, for antenna design. There is a learning curve, but the degree of expertise needed to design a workable antenna for a particular scenario is much reduced with these toolkits. They will produce a CAD file that can be sent to a vendor for manufacturing. This opens up many possibilities for collection opportunities.

Portland offered beautiful surroundings to explore during the week.

Photo courtesy of AP-S