Analysis on Multi-robot Relative 6-DOF Pose Estimation Error Based on UWB Range

Xinran Li, Shuaikang Zheng, Pengcheng Zheng, Haifeng Zhang, Zhitian Li, Xudong ZouarXiv 2023 · DOI · arXiv

Introduction

In environments where GPS/GNSS signals are unavailable—such as dense forests, indoor warehouses, or deep urban canyons—robots must rely on alternative methods for relative positioning. Ultra-Wideband (UWB) ranging has emerged as a leading solution due to its low power consumption and high precision. However, the accuracy of relative 6-DOF (six degrees of freedom) pose estimation is highly dependent on the physical configuration of the UWB tags and anchors on the robots. Poor placement can lead to "degeneracy," where the system cannot distinguish between different orientations, resulting in significant estimation errors.

This paper provides a rigorous mathematical framework to understand how tag geometry affects pose estimation. By deriving an analytical lower bound for error, the authors offer concrete guidance for designing robust multi-robot systems.

Methodology

The research employs a combination of geometric analysis and estimation theory to quantify error bounds. The core of the approach involves analyzing the Jacobian matrix of the measurement model. The smallest eigenvalue of this matrix represents the "weakest link" in the pose estimation, indicating the direction in which the system is most sensitive to noise.

Geometric relationship between the host robot (anchor frame FG) and the tracked robot (tag frame FB).

The authors utilize the Cramer-Rao Lower Bound (CRLB) to establish the theoretical minimum error achievable by any unbiased estimator. They define a "tag simplex"—the geometric shape formed by the arrangement of tags on a robot. Through mathematical proofs, they demonstrate that for a given volume or circumradius, a regular tetrahedron (for 3D space) or a regular triangle (for 2D/4-DOF motion) is the optimal configuration for minimizing error. To solve the actual coordinate transformation between robots, the Levenberg-Marquardt (L-M) non-linear least squares method is applied.

Key Insights: The "Shortest Altitude" Rule

The study identifies the "shortest altitude" of the tag simplex as the most critical geometric factor limiting accuracy. If the tags are placed in a nearly flat configuration, the orientation error increases dramatically.

Key theoretical findings include:

  • Linear Predictability: When the distance between robots ($d$) is significantly larger than the dimensions of the robots themselves, the estimation error behaves linearly. Specifically, the error is proportional to the distance $d$ and inversely proportional to the altitude of the tag simplex ($z_a$ or $h_1$).
  • Minimum Configuration Requirements: To avoid pose ambiguity in 3D, at least four non-coplanar tags/anchors are required. For 4-DOF motion (translation plus heading), two tags may suffice if they are not parallel to the axis of rotation.
  • Euclidean Invariance: The singular values of the Jacobian matrix are invariant under Euclidean transformations, meaning these geometric rules apply regardless of the global coordinate system used.

Key Results and Validation

The authors validated their theoretical findings through extensive Matlab-based Monte-Carlo simulations, using a standard deviation of 5 cm for ranging noise. The results showed a remarkably high Pearson correlation (up to 0.9994) between the proposed linear analytical bounds and the actual simulated Root-Mean-Square Error (RMSE).

RMSE of translation as a function of distance (d) and the inverse of the tag simplex altitude (1/za).

As shown in the simulation results, translation error is primarily driven by the distance between robots and the geometry of the anchor tetrahedron, while orientation error is highly sensitive to the specific "tag simplex" geometry. The maximum orientation error can be reliably approximated by the formula $C_2\sigma_d \frac{d}{z_a h_1} + D_2$, where $\sigma_d$ is the ranging noise.

Conclusion and Future Work

This work establishes that the geometric configuration of UWB tags is not just a secondary concern but a primary determinant of system performance. By following the "Shortest Altitude" rule and utilizing regular simplex configurations, engineers can significantly improve the reliability of multi-robot coordination.

Future research will look to expand this static geometric analysis by incorporating UWB-IMU fusion. Integrating inertial sensors and timing analysis will help researchers understand how motion dynamics and sensor synchronization affect accuracy during high-speed maneuvers or complex docking procedures.

Summary updated

BibTeX
@misc{li2023analysis,
  author = {Xinran Li and Shuaikang Zheng and Pengcheng Zheng and Haifeng Zhang and Zhitian Li and Xudong Zou},
  title = {{Analysis on Multi-robot Relative 6-DOF Pose Estimation Error Based on UWB Range}},
  year = {2023},
  eprint = {2309.15367},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2309.15367},
}
Figure