Design of an adapted extended Kalman filter for enhancing UAV navigation with optical flow

In this paper, the modified design of the currently common visual relative motion inertial navigation system (INS) is proposed to enhance the inertial measurement unit (IMU) model algorithm. The aim of the study is to improve the precise landing and indoor positioning-free flight of existing indoor drones. Both the INS and the airborne IMU used for traditional optical flow guidance are designed to input information to the same extended Kalman filter (EKF) functional module, low cost but resulting in signal phase shift and reduced reliability indicators. Unlike large drones, which can easily reduce vibration, for micro drones, the camera usually must be fixed directly to the body, making it very difficult to isolate vibrations. Using traditional methods can easily cause optical flow guidance to drag down the quality of the entire positioning function. In this study, the improved positioning guidance procedure is proposed and is applied to the indoor positioning flight of UAVs. The EKF, based on the kinematics and optical flow characteristic states equations is separated from the main navigation system. The current positioning and heading are predicted in a way where the optical flow INS and IMU are positioned separately. The positioning and heading are estimated through the current trust weighting value of the optical flow INS and the IMU. Experimental results show that the Monte Carlo method is used to simulate the impact of different signal-to-noise ratios (SNR) on acceleration, angular velocity measurement noise, and the number of detected features on the probability of success of the estimation process. At lower SNR, it shows that the proposed method can resist interference more effectively.