Mobility-Robust Beamforming for Multi-Cell Integrated Sensing and Communication Using Temporal Federated Learning

Abstract

The development of next-generation wireless technologies heavily relies on Integrated sensing and communication (ISAC) for its ability to handle both data transfer and environmental awareness using a singular infrastructure, leveraging shared spectrum or hardware. However, multi-cell ISAC systems face issues with inter-cell interference, heterogeneous statistics from cell to cell, and user mobility on the time variance of the channel itself. Coordination of beamforming has reached a deadlock as federated learning (FL) and personalized-FL techniques only apply to the instantaneous channel state or snapshot of the channel. As a result of this process, FL techniques fail to account for what has been experienced through the channel over a long period of time. Due to this, FL and personalized FL techniques are unable to be applied to high-mobility regimes. In this paper, we propose a mobility aware Beamforming Temporal Federated Learning (TFL) framework for a Integrated Sensing and Communication (ISAC) multi-cell system. We introduce a novel method to leverage temporal sequence of prior communication/sensing channels as input to the federated learning framework allowing for adaptive and robust beamforming to the user's mobility and movement of sensing targets and Doppler effect. Experimental findings indicate that TFL achieves quicker convergence and delivers a superior weighted ISAC utility for users, alongside improved communication and sensing performance, and a more potent blend of communication-sensing advantages compared to per-cell learning, standard federated learning, privacy-compromising federated learning, and mobility-aware federated learning approaches.