Manifold-Optimized Error-State Kalman Filter for Robust Pose Estimation in Unmanned Aerial Vehicles

• DOI: 10.26689/jera.v9i2.10093
• Publication Date: 2025-04-24T01:18:19Z
• Authors: Bolin Jia; Zongwen Bai; Yiqun Gao; Dong Wang; Meili Zhou; Peiqi Gao; Pei Zhang; Zhang Yang

This paper presents a manifold-optimized Error-State Kalman Filter (ESKF) framework for unmanned aerial vehicle (UAV) pose estimation, integrating Inertial Measurement Unit (IMU) data with GPS or LiDAR to enhance estimation accuracy and robustness. We employ manifold-based optimization approach, leveraging exponential logarithmic mappings transform rotation vectors into matrices. The proposed ESKF ensures state variables remain near the origin, effectively mitigating singularity issues enhancing numerical stability. Additionally, due small magnitude of variables, second-order terms can be neglected, simplifying Jacobian matrix computation improving computational efficiency. Furthermore, we introduce novel filter gain strategy that dynamically adapts low-dimensional high-dimensional observation equations, enabling efficient processing across different sensor modalities. Specifically, resource-constrained UAV platforms, this method significantly reduces cost, making it highly suitable real-time applications.