A Nonlinear Double Model for Multisensor-Integrated Navigation Using the Federated EKF Algorithm for Small UAVs.

Area of Science:

• Aerospace Engineering
• Robotics
• Navigation Systems

Background:

• Centralized Kalman filters in integrated navigation suffer from poor robustness and accuracy.
• Small unmanned aerial vehicles (UAVs) require robust and precise navigation, especially for near-ground, short-distance flights.
• Existing multisensor navigation techniques may not adequately address nonlinearities and specific UAV requirements.

Purpose of the Study:

• To propose an improved decentralized federated extended Kalman filter (EKF) for integrated navigation in UAVs.
• To enhance the robustness and solution accuracy of navigation systems compared to traditional methods.
• To address the specific navigation needs of UAVs, particularly attitude solution accuracy.

Main Methods:

• Developed a nonlinear double model incorporating Attitude Heading Reference System (AHRS) and Strapdown Inertial Navigation System (SINS)/GPS models.
• Implemented an improved decentralized federated EKF for multisensor-integrated navigation, optimizing common state parameters.
• Established and simplified a multisensor error model tailored for near-ground, short-distance UAV applications.

Main Results:

• The proposed algorithm demonstrated superior robustness and state estimation accuracy in simulations and flight tests.
• Comparison with Multisensor Complementary Filtering (MSCF) and Multisensor EKF (MSEKF) showed significant improvements.
• The nonlinear double model effectively optimized attitude estimation for UAVs.

Conclusions:

• The improved decentralized federated EKF offers a robust and accurate solution for integrated navigation in small UAVs.
• The nonlinear double model approach effectively enhances attitude determination accuracy.
• This method provides a viable alternative to existing navigation algorithms for UAV applications.