A Novel V2V Cooperative Collision Warning System Using UWB/DR for Intelligent Vehicles

Mingyang Wang, Xinbo Chen, Baobao Jin, Pengyuan Lv, Wei Wang, Yong Shen2021 · DOI

TL;DR

This paper introduces a novel Vehicle-to-Vehicle (V2V) Cooperative Collision Warning System (CCWS) that leverages Ultra-Wideband (UWB) for precise relative positioning and Dead Reckoning (DR) for enhanced accuracy. By fusing UWB and DR data using an Extended Kalman Filter (EKF), the system achieves centimeter-level accuracy in relative positioning and orientation, significantly improving upon traditional methods and enabling more reliable collision warnings.

What Problem Does This Paper Address?

Road safety remains a critical concern, with a significant percentage of accidents attributed to delayed driver responses, particularly rear-end collisions. Existing collision warning systems (CWS) often rely on sensors like radar, lidar, or cameras, which can be hampered by adverse weather conditions and have limitations in accurately capturing a vehicle's relative attitude and motion trends. While Vehicle-to-Vehicle (V2V) communication offers a promising avenue for cooperative systems, current approaches often implement relative positioning and communication separately, potentially impacting real-time performance. This paper addresses the need for a robust, accurate, and real-time V2V CCWS that can overcome the limitations of individual sensors and existing communication methods.

Methods

The proposed CCWS integrates three core components:

  • UWB-based Relative Positioning and Directing System: This system utilizes four UWB modules strategically placed on each vehicle's body corners. An "over-constrained localization method" is employed to enhance the accuracy of calculating the relative position and orientation between vehicles. This method leverages multiple UWB measurements and an optimization approach to find the most accurate solution, even in the presence of ranging errors.
  • Dead Reckoning (DR) System: This subsystem uses wheel-speed sensors on the rear wheels of each vehicle to measure their velocities and yaw rates. This data is crucial for estimating the vehicle's motion and orientation over time.
  • EKF-based Fusion Algorithm: An Extended Kalman Filter (EKF) is used to fuse the data from the UWB and DR systems. This fusion process combines the high-precision, short-range measurements from UWB with the continuous motion estimation from DR, leading to improved accuracy and stability in estimating the relative position, orientation, and velocities of the vehicles.
  • Time to Collision (TTC) Estimation: The system estimates the Time to Collision (TTC) based on the predicted collision position derived from the fused UWB/DR data. This allows for timely warnings to drivers.

Datasets

The paper does not explicitly mention the use of pre-existing datasets for training. Instead, it relies on:

  • Simulated Driving Scenarios: The authors generated various driving scenarios in MATLAB to evaluate the performance of their proposed system, including straight driving and curved routes.
  • Experimental Data: Real-world experiments were conducted using two vehicles equipped with the proposed UWB/DR system. The results were compared against a commercial millimeter-wave radar (MMWR) and data from a high-precision integrated positioning system (RTK-GPS/INS).

Key Insights and Ideas

  • UWB for Relative Kinematics: The paper demonstrates the effectiveness of UWB not just for positioning but also for accurately determining relative orientation between vehicles, which is crucial for advanced collision warning.
  • Over-Constrained Localization for Robustness: The proposed over-constrained method for UWB localization significantly improves accuracy by leveraging redundant measurements and an optimization approach, making it more resilient to ranging errors.
  • Synergistic Fusion of UWB and DR: The EKF fusion of UWB and DR data proves highly effective, enhancing both the accuracy and stability of relative positioning, orientation, and velocity estimation beyond what either system can achieve alone.
  • Standardized Testing for Validation: The system's performance was validated against the JT/T883-2014 standard for forward collision warning systems, demonstrating its practical applicability and compliance with industry requirements.
  • Superiority in Curved Scenarios: The experimental results highlight the proposed system's advantage over a commercial MMWR, particularly in curved driving scenarios, where it shows improved accuracy in relative distance and azimuth.

Key Results

  • Positioning and Directing Accuracy: The UWB-based algorithm with the distance constraint (d5) significantly improved the accuracy of relative longitudinal position (x) and orientation (β) compared to the method without the constraint. RMSE for x improved from 0.70m to 0.21m, and for β from 46.29° to 2.42°.
  • UWB/DR Fusion Performance: The EKF fusion of UWB and DR data further enhanced accuracy. RMSE for relative longitudinal position (x) was reduced to 0.06m, relative lateral position (y) to 0.17m, and relative orientation (β) to 0.83°.
  • Yaw Rate and Velocity Estimation: The UWB/DR fusion system also significantly improved the accuracy of yaw rate and velocity estimations compared to DR alone. For example, RMSE for yaw rate of vehicle 1 decreased from 8.92°/s to 4.60°/s.
  • Collision Warning Success Rate: In simulated rear-end collision scenarios, the system achieved a high "Correct" evaluation rate of 194 out of 196 cases. Across a broader range of random scenarios, the collision warning success rate was 97.9%.
  • Experimental Validation: The system passed all tests according to the JT/T883-2014 standard for straight driving experiments. In curved driving experiments, the proposed system (with fusion) demonstrated superior accuracy compared to a commercial MMWR in middle-distance mode (44% better distance accuracy, 69% better azimuth accuracy) and comparable velocity accuracy.

Limitations and Future Work

  • Velocity Accuracy: The authors acknowledge that the velocity accuracy of the proposed system, while comparable to MMWR, can be further improved. They attribute this to the DR system's reliance on a theoretical Ackerman steering model that doesn't account for real-world factors like suspension stiffness, tire slip, and side-slip angles.
  • Future Directions: The authors suggest future research should focus on developing a system with a more accurate vehicle dynamic model and integrating additional sensors like IMU and GPS to further enhance performance.

Summary updated

BibTeX
@misc{wang2021novel,
  author = {Mingyang Wang and Xinbo Chen and Baobao Jin and Pengyuan Lv and Wei Wang and Yong Shen},
  title = {{A Novel V2V Cooperative Collision Warning System Using UWB/DR for Intelligent Vehicles}},
  year = {2021},
  doi = {10.3390/s21103485},
}
Figure