Multi-Modal Tightly Coupled Robust Pose Estimation for Mobile Robots in Complex Degraded Scenarios

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Multi-modal pose estimation methods face challenges with localization divergence and computational cost in extreme scenarios.
• Sudden illumination variations, geometric degeneracy, and wheel slippage degrade performance of existing algorithms.

Purpose of the Study:

• To develop a tightly coupled multi-modal pose estimation algorithm for mobile robots.
• To enhance robustness and real-time efficiency in challenging environments.

Main Methods:

• Utilizes adaptive robust manifold filtering.
• Employs a pre-integration-driven iterated error-state Kalman filter (iESKF) on a Lie group manifold.
• Integrates Mahalanobis distance chi-square test and M-estimation for adaptive noise isolation.
• Incorporates a perception health quantification system and a smooth degradation state machine.

Main Results:

• Achieves an average processing time of 12.8 ms per frame on edge computing platforms.
• Limits end-to-end closed-loop drift to 1.37 m over a 100-m trajectory in degraded environments.
• Reports a relative translation error (RTE) of approximately 1.2% to 1.4%.

Conclusions:

• The proposed algorithm offers a superior balance between real-time efficiency and robust survivability.
• Demonstrates effective performance in extreme and composite environmental conditions for mobile robot pose estimation.