Channel selection of hybrid communication systems using machine learning

Commercial and military free space communication needs are growing at a remarkable pace. The requirements for free space data transfer have increased as the technology increased, and now engineers are considering higher frequency bands for free space, including W/V, millimeter wave and optical bands. Laser communications can offer much higher data rates than traditional radio frequency (RF) systems and have the added advantage of not being regulated by the International Telecommunication Union (ITU). This along with typical hardware being much smaller in size, weight and power (SWaP) make optical communications a desirable solution for high-rate communication systems. W/V band systems also have the advantage of higher data rates and less frequency congestion than traditional RF systems, as well as the advantages of being able to link through clouds. As the need for “data on demand” increases, the likelihood of users moving to systems that can switch between multiple radios (hybrid systems) is very high. Today, multiple users in both commercial and government sectors are looking to integrate both laser communication (lasercom) and RF solutions onto their platforms. The goal of this work is to utilize a pre-trained Alexnet Convolution Neural Net (CNN) on Doppler radar and GOES satellite imagery to make decisions on which a hybrid system will provide the highest performance differing atmospheric conditions. This method can be scaled for any terrestrial or space-based system.