TVWS Geolocation Database for Secondary-User TVWS Devices for Spectrum Forecasting

Abstract

This paper suggests a MATLAB-based television white space (TVWS) database with forecastingcapabilities. The availability of multiple TV frequencies for a secondary user, depending on the day and time ofthe inquiry, location, and device, was forecasted using reinforcement learning (RL) software. A MATLAB livescript RL application was created and tested. After passing numerous tests, the algorithm was integrated into aMATLAB App Designer application so that an SU could request projections for spectrum availability. Threecategories might be used to categorize the forecast—medium-term (MT), long-term (LT), and short-term (LT).The forecast for that day is referred to as ST, the following day as MT, and the days after that as LT. Based onthe SU's device, the program was designed to predict the spectrum availability in relation to the SU's query time,location, and transmission details. The SU device is a fixed white space device with an EIRP power cap of 4watts. The database and AI are both necessary for accurate predictions. The forecast's results indicated 100%accuracy, assuming the database is frequently updated. Both the forecasting and reinforcement learningprograms were created using MATLAB's App Designer.