This paper addresses the integration of sensing and communications (ISAC) in multipath channels, a critical aspect for predictive beamforming in wireless communications. The authors propose a novel ISAC-assisted beam-tracking solution that leverages reflected echoes to measure kinematic parameters and employs extended Kalman filtering (EKF) for angle prediction. The EKF-ISAC method is shown to outperform conventional feedback-based methods, reducing alignment costs and improving accuracy. However, the angle parameters observed from radar echoes are not always optimal, leading to a gap between the achievable rates and the maximum rates. To bridge this gap, a fine beam tracking method is introduced, which further optimizes the alignment direction based on the EKF-ISAC predictions. Simulation results demonstrate the effectiveness of the proposed method, highlighting its superior performance over traditional feedback-based approaches. The study also explores the impact of various factors such as the number of clusters, paths, and antennas on the achievable rates, providing insights into the trade-offs between time efficiency and alignment accuracy.This paper addresses the integration of sensing and communications (ISAC) in multipath channels, a critical aspect for predictive beamforming in wireless communications. The authors propose a novel ISAC-assisted beam-tracking solution that leverages reflected echoes to measure kinematic parameters and employs extended Kalman filtering (EKF) for angle prediction. The EKF-ISAC method is shown to outperform conventional feedback-based methods, reducing alignment costs and improving accuracy. However, the angle parameters observed from radar echoes are not always optimal, leading to a gap between the achievable rates and the maximum rates. To bridge this gap, a fine beam tracking method is introduced, which further optimizes the alignment direction based on the EKF-ISAC predictions. Simulation results demonstrate the effectiveness of the proposed method, highlighting its superior performance over traditional feedback-based approaches. The study also explores the impact of various factors such as the number of clusters, paths, and antennas on the achievable rates, providing insights into the trade-offs between time efficiency and alignment accuracy.