Generic Multisensor Integration Strategy and Innovative Error Analysis for Integrated Navigation

A modern multisensor integrated navigation system applied in most of civilian applications typically consists of GNSS (Global Navigation Satellite System) receivers, IMUs (Inertial Measurement Unit), and/or other sensors, e.g., odometers and cameras. With the increasing availabilities of low-cost sensors, more research and development activities aim to build a cost-effective system without sacrificing navigational performance. Three principal contributions of this dissertation are as follows: i) A multisensor kinematic positioning and navigation system built on Linux Operating System (OS) with Real Time Application Interface (RTAI), York University Multisensor Integrated System (YUMIS), was designed and realized to integrate GNSS receivers, IMUs, and cameras. YUMIS sets a good example of a low-cost yet high-performance multisensor inertial navigation system and lays the ground work in a practical and economic way for the personnel training in following academic researches. ii) A generic multisensor integration strategy (GMIS) was proposed, which features a) the core system model is developed upon the kinematics of a rigid body; b) all sensor measurements are taken as raw measurement in Kalman filter without differentiation. The essential competitive advantages of GMIS over the conventional error-state based strategies are: 1) the influences of the IMU measurement noises on the final navigation solutions are effectively mitigated because of the increased measurement redundancy upon the angular rate and acceleration of a rigid body; 2) The state and measurement vectors in the estimator with GMIS can be easily expanded to fuse multiple inertial sensors and all other types of measurements, e.g., delta positions; 3) one can directly perform error analysis upon both raw sensor data (measurement noise analysis) and virtual zero-mean process noise measurements (process noise analysis) through the corresponding measurement residuals of the individual measurements and the process noise measurements. iii) The a posteriori variance component estimation (VCE) was innovatively accomplished as an advanced analytical tool in the extended Kalman Filter employed by the GMIS, which makes possible the error analysis of the raw IMU measurements for the very first time, together with the individual independent components in the process noise vector.