Fast and Robust Position and Attitude Estimation Method Based on MARG Sensors

With the development of Internet of Things (IoT), people’s demand for location-based services is increasingly urgent. Based on magnetometer, accelerometer, and rate gyro, i.e., MARG sensors, the position and attitude estimation methods such as complementary filter (CF), Kalman filter (KF), and their various modifications have been the research hot spot. However, the CF-based methods are empirical and lack robustness; the KF-based methods are memory-free observers, whose solution may diverge when the filter lacks uniform observability. In this article, a virtual-measurement-combined extended KF (VMC-EKF) method is proposed by fusing the carrier’s motion state with EKF method. Similar to the graph optimization, the proposed method can measure the key information in VMC phase, and thus remove the requirement of uniform observability. The number of virtual measurements can be estimated based on carrier’s motion state and its gradient, which determines the number of iterations of the prediction phase and the correction phase. In order to verify the performance of the proposed method, a series of numerical simulation experiments, turntable experiments, and foot-mounted experiments are carried out. The corresponding testing platform is set up based on MPU9250, which is a typical and low-cost motion tracking integrated circuit (IC) of MARG sensors. The raw data of MARG sensors can communicate with the host computer via wired or wireless communication, and then be imported into MATLAB for processing and analysis by the compared methods. The test results show that the proposed method can achieve fast convergence of attitude estimation and avoid the divergence of position estimation compared with the state-of-the-art methods.