MACHINE LEARNING DRIVEN ADAPTIVE RIS PARTITIONING AND RESOURCE OPTIMIZATION FOR INTEGRATED SENSING AND COMMUNICATION

Abstract

Due to the fast evolution of 6G wireless communication networks, the emergence of the Integrated Sensing and Communication (ISAC) technique becomes necessary because communication and sensing can be done simultaneously by sharing a common wireless resource. This paper presents a novel reconfigurable intelligent surface (RIS) assisted ISAC architecture for the efficient optimization of communication embedded sensing. Intheproposedsystemplan, theallocationofRISelementsforcommunication and sensing operations are dynamically optimized, as well as communication-sensing priority and power allocation, in order to make the overall performance optimal in non stationary wireless environments. In addition, Nakagami-m fading channel models are taken to present real-time features of the communication and sensing conditions. A multi-objective rewardfunction isdevelopedusingthesensingandcommunication rates and thecommunication-sensing tradeoffperformance. VariousarrangementsoftheRIS are analyzed to determine the best configuration of division for communications and sensing depending on different scenarios in terms of wireless channel characteristics. An extensive database of simulations is obtained via exhaustive optimization, and an intelligent predicting algorithm of learning of Random Forest is applied to predict the best RIS configuration comparing it with a KNN Classifier which is based on the extracted communications and sensing metrics. The accuracy of predicting the best RIS dividing configuration using the Random Forest method exceeds those achieved using other conventional techniques. This framework provides a reliable and efficient solution for ISAC-RIS systems which supports wireless networks also beyond 6G.