Vehicle Trajectory Estimation in GNSS denied Environments: An analysis of Extended Kalman Filter Approach for Monocular Vision Aided Inertial Navigation

This paper begins by describing the Extended Kalman Filter (EKF) formulation for vehicle trajectory estimation in GNSS denied environments using tracked static features and their known positions with IMU-CAM-EKF sensor system. The paper demonstrates with simulations that although the EKF approach is computationally efficient, the theoretical analysis of the mathematical models suggests that the both the system as well as the measurement models can be improved. This improvement is not only limited to the system and measurement models but the augmentation of the state vector such as maintaining the history of past states can also contribute towards optimal trajectory estimation. To consolidate the analysis, a scenario is considered for a landing vehicle, whose monocular camera and its associated image processing module can detect some known static features on the ground whose position is known in advance or stored in on-board computer. Then the paper begins with the analysis and improvements of the mathematical models and provide a roadmap for trajectory estimation in unknown environments such as when the position of tracked features is unknown.