Research on UAV Robust Adaptive Positioning Algorithm Based on IMU/GNSS/VO in Complex Scenes.

Area of Science:

• Robotics
• Navigation Systems
• Artificial Intelligence

Background:

• Unmanned Aerial Vehicles (UAVs) require robust autonomous positioning for intelligent systems.
• Challenges arise from Global Navigation Satellite System (GNSS) signal interruption and insufficient visual features in complex environments.
• Existing methods struggle with accuracy and reliability in degraded sensor conditions.

Purpose of the Study:

• To develop a novel robust adaptive positioning algorithm for UAVs.
• To enhance the accuracy and robustness of autonomous navigation and positioning systems.
• To address limitations in current positioning methods under challenging environmental conditions.

Main Methods:

• A multi-source fusion model based on a federated Kalman filter (FKF) was established, combining vision/inertial navigation and satellite/inertial navigation.
• An Extended State Kalman Filter (ESKF) was utilized within the fusion framework.
• A robust adaptive localization algorithm was proposed, employing robust equivalent weights and adaptive information sharing coefficients.

Main Results:

• The proposed algorithm effectively mitigates the impact of gross errors and model deviations.
• It adaptively updates information sharing coefficients based on sub-filter covariance.
• Position, velocity, and attitude measurement accuracy improved by 2-3 times compared to classical and adaptive FKF algorithms.
• Real-time navigation requirements are met.

Conclusions:

• The robust adaptive localization algorithm significantly improves the reliability and accuracy of autonomous UAV navigation in complex scenes.
• The algorithm demonstrates generality, applicable to various sensor combinations and environments.
• It offers a reliable solution for autonomous positioning challenges faced by UAVs.