At all stages of radar development and operation, we rely, implicitly or explicitly, on models designed to encapsulate what we regard as relevant characteristics of the natural world. In order to be of practical use, these models are formulated in the language of mathematics, and nowadays that usually leads to a second level of abstraction, namely, representation in an algorithmic framework amenable to execution in computers. These models are constructed on the foundation of fundamental constraints such as Newton’s laws of motion, Maxwell’s equations, the Navier-Stokes equation, and a law of gravitation, together with empirical information about the constitutive properties of the media involved. If we were to discuss relativistic targets or quantum radar, we would need to add Einstein’s relativity or the Schrodinger and perhaps Dirac equations, but here we will limit our attention to classical radars in terrestrial settings, specifically, radars operating in the HF band. Einstein is said to have remarked that “things should be made as simple as possible – but no simpler” and that would seem to be good advice when selecting models. Unfortunately, several factors intervene to frustrate this approach, leading to the situation where the performance of many operational radars is unnecessarily compromised. First, to paraphrase Eugene Wigner, we must point to the unreasonable effectiveness of linearisation, localisation and Gaussianity. Nature has been extraordinarily kind to us by determining that highly tractable mathematico-physical approximations to the fundamental equations yield PDG (pretty damn good) descriptions of radio wave propagation, scattering and environmental dynamics. If a simple model does the job, why strive to make life more complicated, for possibly marginal gain? Second, the solutions to the approximate equations have easily-understood physical interpretations that are helpful when specifying radar parameters and interpreting radar echoes, whereas more sophisticated models often provide quite abstruse descriptions. Third, situation assessment and parameter estimation based on simple models tend to be mathematically straightforward and computationally undemanding. Fourth, technology continues to advance rapidly, and this generally results in improved radar performance, even when the underlying models are not updated. This phenomenon creates the illusion that the technological advances have borne their fruit. In this talk I propose to argue the case for more advanced models in the context of HF radar, using examples from radar design, signal acquisition, processing and analysis, echo interpretation, and implementation of procedures associated with artificial intelligence. Co-sponsored by: Shengjian Jammy Chen Speaker(s): , Dr. Stuart Anderson Room: N136, Bldg: Engineering North, N136 Culver Room, Adelaide, South Australia, Australia, 5005
