Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Successive Pose Estimation and Beam Tracking for mmWave Vehicular Communication Systems

Cen Liu, Guangxu Zhu, Fan Liu, Yuanwei Liu, Kaibin Huang

TL;DR

The paper tackles high-mobility mmWave vehicular links where beam training overhead is prohibitive, by fusing mmWave radar sensing with communications through a SPEBT framework. It first derives 2D vehicle pose from radar point clouds using Fast-CFEAR, and then uses the pose estimates to drive EKF-based beam tracking for the LoS channel, formulating a pose-estimation–aware channel evolution model. The approach achieves accurate pose and beam tracking while reducing beam training overhead to less than 5% on real-world data, and demonstrates robustness under GNSS-denied and adverse conditions. This has practical impact for reliable mmWave vehicular links in challenging environments, with potential extensions to 3D sensing and tracking.

Abstract

The millimeter wave (mmWave) radar sensing-aided communications in vehicular mobile communication systems is investigated. To alleviate the beam training overhead under high mobility scenarios, a successive pose estimation and beam tracking (SPEBT) scheme is proposed to facilitate mmWave communications with the assistance of mmWave radar sensing. The proposed SPEBT scheme first resorts to a Fast Conservative Filtering for Efficient and Accurate Radar odometry (Fast-CFEAR) approach to estimate the vehicle pose consisting of 2-dimensional position and yaw from radar point clouds collected by mmWave radar sensor. Then, the pose estimation information is fed into an extend Kalman filter to perform beam tracking for the line-of-sight channel. Owing to the intrinsic robustness of mmWave radar sensing, the proposed SPEBT scheme is capable of operating reliably under extreme weather/illumination conditions and large-scale global navigation satellite systems (GNSS)-denied environments. The practical deployment of the SPEBT scheme is verified through rigorous testing on a real-world sensing dataset. Simulation results demonstrate that the proposed SPEBT scheme is capable of providing precise pose estimation information and accurate beam tracking output, while reducing the proportion of beam training overhead to less than 5% averagely.

Successive Pose Estimation and Beam Tracking for mmWave Vehicular Communication Systems

TL;DR

The paper tackles high-mobility mmWave vehicular links where beam training overhead is prohibitive, by fusing mmWave radar sensing with communications through a SPEBT framework. It first derives 2D vehicle pose from radar point clouds using Fast-CFEAR, and then uses the pose estimates to drive EKF-based beam tracking for the LoS channel, formulating a pose-estimation–aware channel evolution model. The approach achieves accurate pose and beam tracking while reducing beam training overhead to less than 5% on real-world data, and demonstrates robustness under GNSS-denied and adverse conditions. This has practical impact for reliable mmWave vehicular links in challenging environments, with potential extensions to 3D sensing and tracking.

Abstract

The millimeter wave (mmWave) radar sensing-aided communications in vehicular mobile communication systems is investigated. To alleviate the beam training overhead under high mobility scenarios, a successive pose estimation and beam tracking (SPEBT) scheme is proposed to facilitate mmWave communications with the assistance of mmWave radar sensing. The proposed SPEBT scheme first resorts to a Fast Conservative Filtering for Efficient and Accurate Radar odometry (Fast-CFEAR) approach to estimate the vehicle pose consisting of 2-dimensional position and yaw from radar point clouds collected by mmWave radar sensor. Then, the pose estimation information is fed into an extend Kalman filter to perform beam tracking for the line-of-sight channel. Owing to the intrinsic robustness of mmWave radar sensing, the proposed SPEBT scheme is capable of operating reliably under extreme weather/illumination conditions and large-scale global navigation satellite systems (GNSS)-denied environments. The practical deployment of the SPEBT scheme is verified through rigorous testing on a real-world sensing dataset. Simulation results demonstrate that the proposed SPEBT scheme is capable of providing precise pose estimation information and accurate beam tracking output, while reducing the proportion of beam training overhead to less than 5% averagely.
Paper Structure (19 sections, 23 equations, 5 figures, 3 tables, 1 algorithm)

This paper contains 19 sections, 23 equations, 5 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. MmWave Vehicular Communications Model
  4. MmWave Vehicular Sensing Model
  5. Successive Pose Estimation and Beam Tracking
  6. Fast-CFEAR Based Pose Estimation
  7. $K$-Strongest Conservative Filtering
  8. Downsamping
  9. Estimating Oriented Surface Points
  10. Point Cloud Registration
  11. EKF Based Beam Tracking
  12. State Evolution Model
  13. Measurement Model
  14. The Proposed SPEBT Scheme
  15. Simulation Results
  16. ...and 4 more sections

Figures (5)

  • Figure 1: mmWave vehicular communications system. Left: 3D view of the MIMO system. Right: 2D bird's eye view of the MIMO system.
  • Figure 2: mmWave vehicular sensing platform Barnes20. Left: a Navtech CTS350-X mmWave FMCW radar (marked with a blue box) mounted on a Nissan LEAF automobile. Middle: a sample of raw radar point cloud in polar form. Right: a sample of raw radar point cloud in Cartesian form.
  • Figure 3: The estimated vehicle moving trajectory compared to ground truth. Both trajectories are aligned by the initial position marked with $\times$, and the final positions are marked with $\square$. The position of BS is marked with $+$.
  • Figure 4: The estimated pose compared to ground truth pose on coordinate $(x_k,y_k)$ and yaw $\theta_k$ against timeslot $k$.
  • Figure 5: A beam tracking realization generated by SPEBT scheme compared to ground truth channel on $|\alpha_k|$, $\phi_k$ and $\psi_k$ against timeslot $k$.