Toward Reliable and Accurate Predictive ISAC in Mobile mmWave Networks

As 6G networks aim to integrate high-precision sensing with high-speed communication, Integrated Sensing and Communications (ISAC) has emerged as a key technology. In mobile millimeter wave (mmWave) networks, maintaining accurate channel state information (CSI) is challenging due to high mobility, blockage, and the prohibitive pilot overhead required for conventional estimation. This paper introduces a low-complexity, model-driven ISAC framework that leverages sensing-derived target parameters to predict channel states and optimize beamforming without relying on extensive uplink feedback.

The researchers develop a geometry-aware channel prediction framework that reconstructs channel state information by combining line-of-sight (LoS) components with dynamically detected non-line-of-sight (NLoS) components. By treating surrounding vehicles as mobile scatterers, the system extracts spatial features—such as angle of arrival, Doppler shift, and range—using an extended Kalman filter (EKF). The authors formulate a non-convex beamforming optimization problem aimed at minimizing the worst-user Cramér-Rao bound (CRB) while satisfying per-user communication rate constraints. To solve this, they employ successive convex approximation (SCA) and Schur-complement-based decomposition. Additionally, they propose a robust non-coherent beamforming design that uses sample averaging to maintain performance despite unknown absolute phase variations.

Numerical evaluations demonstrate that the proposed framework significantly outperforms conventional communication-based localization and tracking schemes. By explicitly modeling NLoS components, the system maintains reliable connectivity even during severe blockage events, where traditional LoS-dependent systems would fail. The robust non-coherent design ensures that the system remains resilient to hardware impairments and phase noise, providing a practical solution for high-frequency, blockage-prone vehicular networks. The framework effectively balances sensing quality and communication throughput, offering a scalable solution for dense traffic scenarios.