Fault-tolerant Sensor Fusion Based on Inertial Measurements and GNSS
MetadataShow full item record
The standard observer for inertial navigation system (INS) have for many years been the extended Kalman filter. Due to extensive research, in recent years, on nonlinear observer applied with low-cost inertial sensors can this possible change.Fault-tolerance are in many applications necessary. In dynamic positioning operations are fault-tolerance required. This thesis dealt with development of a fault-tolerant nonlinear observer for integration of INS and Global Navigation Satellite Systems (GNSS). Furthermore, the observer was applied for dynamic positioning, by developing a simulator to obtain vessel motion and sensor readings. The main focus were on GNSS errors and faults. Based on this were methods used to detect and handle outlier detection, sensor freeze, high variance of GNSS sensors and GNSS bias. Furthermore, a novel GNSS drift detection algorithm, applicable for marine vessel, was developed. Moreover, senor voting and sensor weighting was carried out by developing a voting algorithm. Also a model-based observer was utilized to provide redundant acceleration information to the INS.The chosen INS/GNSS observer proved to be a good basis for the fault-tolerant additions. Outliers, sensor freeze, high variance and bias of the GNSS sensors were detected and handled accordingly. GNSS drift was detected and a possible drive-off situation was prevented. Furthermore, utilizing a model-based observer to obtain redundant acceleration information was shown to be successful.