Machine learning applied to the ray tracing model of the OTH radar

Over-the-horizon skywave radar (OTHR) uses reflections in the Earth's ionosphere to detect targets at very long ranges. To achieve detection, the electromagnetic energy emitted by the radar must follow a specific path with the target as its endpoint. This propagation path is strongly dependent on three factors: the direction of emission, the frequency of the electromagnetic wave and the conditions of the propagation medium, in particular the dynamic and complex characteristics of the ionosphere. From the above, it can be concluded that for this type of radar, the prior determination of the wave propagation path is essential for the correct operation of the system. In this work we are presenting the first results of a computational model based on Machine Learning to predict the propagation path of an electromagnetic wave emitted by an OTHR. A variety of ray tracing models are currently available. These models can range from very simple versions using analytical equations to versions that find the propagation path by solving a system of differential equations using numerical integration methods. The proposed model will provide solutions to the main disadvantages of traditional ray tracing models based on systems of differential equations, such as: Speed of response: once trained, the neural network model is able to respond to new queries almost instantaneously; Adaptive capability: unlike traditional models, neural networks can be retrained or adapted to incorporate new data and scenarios, improving accuracy, and, finally Reduced computing resources: optimised neural network models can run on less demanding hardware than numerical integration methods.