Papers
Topics
Authors
Recent
Search
2000 character limit reached

Unscented Particle Filter for Visual-inertial Navigation using IMU and Landmark Measurements

Published 27 Apr 2025 in cs.RO, cs.SY, and eess.SY | (2504.19318v1)

Abstract: This paper introduces a geometric Quaternion-based Unscented Particle Filter for Visual-Inertial Navigation (QUPF-VIN) specifically designed for a vehicle operating with six degrees of freedom (6 DoF). The proposed QUPF-VIN technique is quaternion-based capturing the inherently nonlinear nature of true navigation kinematics. The filter fuses data from a low-cost inertial measurement unit (IMU) and landmark observations obtained via a vision sensor. The QUPF-VIN is implemented in discrete form to ensure seamless integration with onboard inertial sensing systems. Designed for robustness in GPS-denied environments, the proposed method has been validated through experiments with real-world dataset involving an unmanned aerial vehicle (UAV) equipped with a 6-axis IMU and a stereo camera, operating with 6 DoF. The numerical results demonstrate that the QUPF-VIN provides superior tracking accuracy compared to ground truth data. Additionally, a comparative analysis with a standard Kalman filter-based navigation technique further highlights the enhanced performance of the QUPF-VIN.

Summary

Unscented Particle Filter for Visual-Inertial Navigation using IMU and Landmark Measurements

The paper presents a novel approach to visual-inertial navigation utilizing an Unscented Particle Filter (UPF) designed for a six degrees of freedom (6 DoF) vehicle, particularly focusing on operations within GPS-denied environments. Navigation challenges, common when GPS signals are unreliable, demand robust algorithms capable of leveraging onboard sensors for position and orientation estimation. Here, the authors propose the Quaternion-based Unscented Particle Filter for Visual-Inertial Navigation (QUPF-VIN), which captures the nonlinear nature of navigation kinematics through quaternion formulation.

Methodology Overview

The core innovation lies in the use of quaternions within the UPF to address nonlinearities inherent in navigation kinematics. By integrating an inertial measurement unit (IMU) along with vision-based landmark measurements, QUPF-VIN effectively fuses data to enhance navigation accuracy. Key to the filter’s design is its compatibility with discrete-time implementation, aligning with the operational metrics of onboard inertial systems. This ensures seamless integration and application to real-world navigation scenarios involving unmanned aerial vehicles (UAVs) equipped with 6-axis IMUs and stereo cameras.

Experimental Validation and Results

The QUPF-VIN approach was validated using the EuRoC micro aerial vehicle dataset, showcasing its efficacy in real-world conditions. The experiments employed a quadrotor equipped with low-cost sensor arrays—stereo cameras and IMUs—illustrating the capability of QUPF-VIN in maintaining tracking accuracy compared to ground truth data. The numerical results indicate significant improvements in position and orientation estimates compared to traditional Kalman filter counterparts. The proposed filter demonstrated rapid convergence in estimation errors across orientation, position, and velocity, emphasizing its robustness and reliability.

Comparative Analysis

In comparison to Kalman Filter-based techniques, including Extended Kalman Filters (EKF) and Multiplicative EKF, QUPF-VIN showed superior handling of system nonlinearities without necessitating linearization around nominal points. Moreover, its ability to capture arbitrary distributions highlights its robustness over conventional UKF approaches, which may fail in scenarios involving narrow sensor distribution profiles.

Implications and Future Directions

The results suggest promising practical applications, particularly for UAVs operating in environments where GPS signals are compromised. The ability of QUPF-VIN to integrate low-cost sensors into robust navigation frameworks opens avenues for further advancements in autonomous navigation systems, potentially enhancing both ground and aerial robotics. Extending this methodology to larger networks of interconnected vehicles could also prove valuable, specifically in swarm robotics and collaborative navigation tasks.

In conclusion, the QUPF-VIN algorithm represents a significant stride in autonomous navigation, efficiently bridging the gap between nonlinearity management and computational feasibility. Future research could explore optimization techniques for real-time applications, leveraging advancements in AI for adaptive navigation strategies. Furthermore, integrating other sensor modalities into the QUPF-VIN framework could facilitate comprehensive navigation solutions across diverse operational domains.

Paper to Video (Beta)

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.

Collections

Sign up for free to add this paper to one or more collections.