The paper discusses the development of A-KIT, an adaptive Kalman-informed transformer, designed to improve the accuracy of sensor fusion in navigation applications. Traditional methods, such as the extended Kalman filter (EKF), often assume constant process noise, leading to inaccuracies in real-world scenarios where noise varies. A-KIT addresses this by learning to adapt to varying process noise covariance online. The paper demonstrates that A-KIT outperforms the conventional EKF by over 49.5% and model-based adaptive EKF by an average of 35.4% in terms of position accuracy.
Publication date: 19 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.09987
