Abstract:In this document, we revise the results of [1] based on more reasonable assumptions regarding data shuffling and parameter setup of deep neural networks (DNNs). Thus, the simulation results can now more reasonably demonstrate the performance of both the proposed and compared beam alignment methods. We revise the simulation steps and make moderate modifications to the design of the vehicle distribution feature (VDF) for the proposed vision based beam alignment when the MS location is available (VBALA). Specifically, we replace the 2D grids of the VDF with 3D grids and utilize the vehicle locations to expand the dimensions of the VDF. Then, we revise the simulation results of Fig. 11, Fig. 12, Fig. 13, Fig. 14, and Fig. 15 in [1] to reaffirm the validity of the conclusions.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to correct and improve the simulation results and designs of the millimeter - wave (mmWave) beam alignment methods in previous studies, especially the vision - assisted beam alignment method (VBALA) in vehicle - to - everything (V2X) communications. Specifically, the authors have made improvements in the following aspects: 1. **Data Shuffling and Parameter Settings**: In the original paper, the dataset was shuffled only once at the initialization during the training phase, which led to an unreasonable performance evaluation. The new paper modifies the training method to ensure that the dataset is shuffled in every epoch and adjusts some parameter settings of the deep neural network (DNN) to more reasonably demonstrate the performance of each method. 2. **Design of Vehicle Distribution Feature (VDF)**: The original paper used a two - dimensional - grid - based VDF. The new paper replaces it with a three - dimensional grid and extends the dimension of the VDF by using vehicle positions. Specifically, the cube RSU coverage area is divided into a 3D grid of \( GX\times GY\times GZ \), a local coordinate system (LCS) is set for each grid, and the average length, width, height, azimuth angle, and center coordinates of the vehicle are calculated according to its position. 3. **Correction of Simulation Results**: Multiple key charts in the original paper (such as Figure 11, Figure 12, Figure 13, Figure 14, and Figure 15) are regenerated to verify the validity of the conclusions. Through these corrections, the authors have found some differences from the original conclusions: - Under non - line - of - sight (NLOS) conditions, the performance degradation trend of multi - modal beam alignment (BMBA) is lower than that of VBALA. Therefore, BMBA can outperform VBALA at larger position errors. - Under line - of - sight (LOS) conditions, the performance of VBALA is slightly lower than that of BMBA, but it has advantages in computational complexity and environmental generalization ability. 4. **Supplementary Corrections**: Figure 15 in the original paper wrongly shows the verification accuracy of the predicted beam coherence time (VPBCT), and the new paper corrects this. In general, by improving data processing methods, optimizing feature designs, and re - evaluating simulation results, this paper aims to provide a more accurate and reasonable performance evaluation of millimeter - wave beam alignment, especially in V2X communication scenarios.

Corrections to "Computer Vision Aided mmWave Beam Alignment in V2X Communications"

Beam Alignment in MmWave V2X Communications: A Survey

Joint Communication and Control for mmWave/THz Beam Alignment in V2X Networks

Beamwidth Optimization for Millimeter-Wave V2V Communication Between Neighbor Vehicles in Highway Scenarios

Beam Design For V2v Communications With Inaccurate Positioning Based On Millimeter Wave

Deep learning-based beam alignment in mmWave vehicular networks

Vehicle Cameras Guide mmWave Beams: Approach and Real-World V2V Demonstration

Beam management optimization for V2V communications based on deep reinforcement learning

Integrated Sensing and Communications Towards Proactive Beamforming in mmWave V2I via Multi-Modal Feature Fusion (MMFF)

Intelligent Vehicle Network Routing With Adaptive 3D Beam Alignment for mmWave 5G-Based V2X Communications

Fast Channel Estimation and Beam Tracking for Millimeter Wave Vehicular Communications

Third-party Camera Aided Beam Alignment Real-world Prototype for Mmwave Communications

Position Aware 60 GHz mmWave Beamforming for V2V Communications Utilizing Deep Learning

Location-Assisted Beam Alignment for Train-to-Train Communication in Urban Rail Transit System.

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

Beamforming and Scalable Image Processing in Vehicle-to-Vehicle Networks

Radar Aided mmWave Vehicle-to-InfrastructureLink Configuration Using Deep Learning

Predictive Beam Tracking with Cooperative Sensing for Vehicle-to-Infrastructure Communications.

Deep Learning for Hierarchical Beam Alignment in mmWave Communication Systems

Sensing-Assisted Beam Tracking in V2I Networks: Extended Target Case

Beam Switching Techniques for Millimeter Wave Vehicle to Infrastructure Communications