A Hybrid Positioning Strategy for Vehicles in a Tunnel Based on RFID and In-Vehicle Sensors

Xiang Song, Xu Li, Wencheng Tang, Weigong Zhang, Bin Li2014 · DOI

TL;DR

This paper tackles the challenge of accurate vehicle positioning in GPS-denied environments like tunnels by proposing a hybrid multi-sensor fusion strategy. It combines RFID for preliminary positioning with in-vehicle sensors (IMU, wheel speed, compass) using an Interacting Multiple Model (IMM) framework with a Strong Tracking Extended Kalman Filter (STEKF) for robust and precise localization.

What Problem Does This Paper Address?

Intelligent transportation systems (ITS) rely heavily on accurate and continuous vehicle positioning. While GPS is the go-to technology, it fails in environments like tunnels due to signal blockage. This poses a significant challenge for applications ranging from navigation to safety. Traditional solutions often involve fusing GPS with Inertial Navigation Systems (INS) or Dead Reckoning (DR). However, low-cost MEMS-based inertial sensors used in automotive applications accumulate errors over time, leading to performance degradation during GPS outages, especially in long tunnels. Other sensors can partially correct these errors, but they struggle to compensate for prolonged GPS failures. This paper addresses the need for a reliable and accurate positioning solution specifically for tunnels, where GPS is unavailable.

Methods

The proposed strategy is a two-stage approach:

  • RFID-based Preliminary Positioning: This stage utilizes Radio Frequency Identification (RFID) technology to establish an initial position estimate. The Received Signal Strength (RSS) from RFID tags is used to estimate the distance to the reader. A decentralized Least Mean Square (LMS) federated filter is employed to process these distance measurements and provide preliminary position information. This federated structure offers fault tolerance and high precision.
  • IMM-STEKF-based Global Fusion: To enhance the preliminary positioning, a global fusion algorithm is developed. This algorithm integrates the RFID-based position estimates with data from low-cost in-vehicle sensors, including electronic compasses and wheel speed sensors. The core of this stage is an Interacting Multiple Model (IMM) estimator. The IMM approach allows the system to adapt to different driving conditions by employing multiple models (Constant Acceleration - CA, and Constant Turn - CT) to represent vehicle motion. Crucially, the conventional Extended Kalman Filter (EKF) within the IMM is replaced by a Strong Tracking Extended Kalman Filter (STEKF). STEKF offers improved robustness to system parameter changes, better tracking during dynamic maneuvers, and lower sensitivity to noise compared to EKF.

Datasets

Not discussed in the paper. The experiments were conducted using real-world data collected from a Buick Sail SRV vehicle equipped with various sensors.

Key Insights and Ideas

  • RFID as a Tunnel Anchor: RFID, despite its limitations (low output frequency, no speed/attitude information), can serve as a valuable "anchor" for preliminary positioning in tunnels, providing essential position observations where GPS is absent.
  • Federated Filter for Robustness: The use of a federated LMS filter for RFID-based positioning effectively isolates false information from tags, improving overall reliability.
  • IMM for Dynamic Driving: The IMM approach is critical for handling the diverse driving scenarios within a tunnel, allowing the system to switch between different motion models (straight line vs. turning).
  • STEKF for Enhanced Kalman Filtering: Replacing EKF with STEKF within the IMM framework significantly improves the filter's ability to handle uncertainties in system models and noise characteristics, leading to more stable and accurate estimates.
  • Synergy of RFID and In-Vehicle Sensors: The fusion of RFID's positional data with the richer state information (velocity, heading) from in-vehicle sensors creates a powerful combination that overcomes the individual limitations of each sensor type.

Key Results

  • The RFID-based preliminary positioning algorithm achieved an RMS Euclidean distance error of 3.01 m in the tunnel, a 47% improvement over a standard multilateration method.
  • The proposed IMM-STEKF-based fusion algorithm significantly improved positioning accuracy. In the tunnel, the RMS Euclidean distance error was reduced to 1.73 m, representing a 39% improvement over the RFID-based preliminary positioning.
  • The fusion algorithm increased the positioning frequency to 10 Hz from 1 Hz (RFID-only), and provided velocity and heading angle information, which RFID alone cannot.
  • In comprehensive tests, the IMM method achieved the lowest RMS Euclidean distance error (1.45 m) among all tested methods (RFID, DR, EKF, STEKF), demonstrating superior accuracy and reliability.

Limitations and Future Work

  • Experimental Environment: The experiments were conducted in an outdoor simulated tunnel environment rather than a real enclosed tunnel. The authors acknowledge that multipath phenomena in enclosed environments could affect positioning performance.
  • Tag Density: The performance of the RFID system is dependent on the density of tags. The authors suggest future work will focus on reducing the number of tags needed while minimizing accuracy sacrifice, likely to cut costs.
  • Real Enclosed Environments: Future work will involve addressing the challenges posed by multipath effects in actual enclosed tunnel environments.
  • Cost Reduction: Further research will explore methods to reduce the number of RFID tags required without significantly compromising accuracy.

Summary updated

BibTeX
@misc{song2014hybrid,
  author = {Xiang Song and Xu Li and Wencheng Tang and Weigong Zhang and Bin Li},
  title = {{A Hybrid Positioning Strategy for Vehicles in a Tunnel Based on RFID and In-Vehicle Sensors}},
  year = {2014},
  doi = {10.3390/s141223095},
}
Figure