Abstract:A near-field sensing (NISE) enabled predictive beamforming framework is proposed to facilitate wireless communications with high-mobility channels. Unlike conventional far-field sensing, which only captures the angle and the radial velocity of the user, NISE enables the estimation of the full motion state, including additional distance and transverse velocity information. Two full-motion state sensing approaches are proposed based on the concepts of estimation and tracking, respectively. 1)AGD-AO approach: The full motion state of the user is estimated within a single CPI. In particular, the gradient descent is adopted to estimate the transverse and radial velocities of the user based on the maximum likelihood criteria, while the distance and the angle are calculated by the kinematic model. In this process, moment estimations are leveraged to adaptively tune the step size, thereby leading to a smoother and faster gradient descent. 2) EKF approach: The full motion state of the user is tracked across multiple CPIs. Based on the noisy measurements in multiple CPIs, the EKF iteratively predicts and updates the current motion state to achieve a low tracking error. Based on the obtained full motion state, the beam prediction, and Doppler frequency compensation can be carried out with minimum pilot overhead. Numerical results are provided to validate the effectiveness and efficiency of the proposed approach compared to the conventional far-field predictive beamforming and feedback-based approaches. It is also revealed that: 1)the proposed AGD-AO can achieve stable descending with small gradients, thereby accelerating convergence; 2) compared to far-field predictive beamforming and feedback-based schemes, both of the proposed methods exhibit superior performance; and 3) by incorporating multiple CPIs, the EKF method exhibits greater robustness in low SNR regions.

Beam wander prediction with recurrent neural networks

Recognizing Fractional Orbital Angular Momentum Using Feed Forward Neural Network

Enhancing Mmwave Beam Prediction Through Deep Learning with Sub-6 GHz Channel Estimate

Capacities for Long-Distance Free-Space Optical Links Under Beam Wander Effects

Best Beam Prediction in Non-Standalone Mm Wave Systems

Vortex Beams and Deep Learning for Optical Wireless Communication through Turbulent and Diffuse Media

Machine Learning Prediction for Phase-less Millimeter-Wave Beam Tracking

Joint Optimization of Data- and Model-Driven Probing Beams and Beam Predictor

Camera Based mmWave Beam Prediction: Towards Multi-Candidate Real-World Scenarios

Anticipating Optical Availability in Hybrid RF/FSO Links Using RF Beacons and Deep Learning

Channel Charting Based Beam SNR Prediction

Deep Learning Assisted Mmwave Beam Prediction with Prior Low-frequency Information

Beam prediction and tracking mechanism with enhanced LSTM for mmWave aerial base station

Beam Alignment for Millimetre Wave Links with Motion Prediction of Autonomous Vehicles

Vision-Aided Beam Tracking: Explore the Proper Use of Camera Images with Deep Learning

Machine learning-based adaptive optics for free-space optical communication: a training data generation study

Vision-Position Multi-Modal Beam Prediction Using Real Millimeter Wave Datasets

Near-Field Sensing Enabled Predictive Beamforming: From Estimation to Tracking

Multimodal Deep Learning Empowered Millimeter-Wave Beam Prediction

Angle-of-Arrival Estimation of Narrow Gaussian Beams for Mobile FSO Platforms

Deep-learning-assisted communication capacity enhancement by non-orthogonal state recognition of structured light