Fusion of Redundant Aided-inertial Sensors with Decentralised Kalman Filter for Autonomous Underwater Vehicle Navigation

Abstract

Most submarines carry more than one set of inertial navigation system (INS) for redundancy and reliability. Apart from INS systems, the submarine carries other sensors that provide different navigation information. A major challenge is to combine these sensors and INS estimates in an optimal and robust manner for navigation. This issue has been addressed by Farrell 1 . The same approach is used in this paper to combine different sensor measurements along with INS system. However, since more than one INS system is available onboard, it would be better to use multiple INS systems at the same time to obtain a better estimate of states and to provide autonomy in the event of failure of one INS system. This would require us to combine the estimates obtained from local filters (one set of INS system integrated with external sensors), in some optimal way to provide a global estimate. Individual sensor and IMU measurements cannot be accessed in this scenario. Also, autonomous operation requires no sharing of information among local filters. Hence a decentralised Kalman filter approach is considered for combining the estimates of local filters to give a global estimate. This estimate would not be optimal, however. A better optimal estimate can be obtained by accessing individual measurements and augmenting the state vector in Kalman filter, but in that case, corruption of one INS system will lead to failure of the whole filter. Hence to ensure satisfactory performance of the filter even in the event of failure of some INS system, a decentralised Kalman filtering approach is considered.

Low Cost Integrated Navigation System for Unmanned Vessel

“…INS and DVS are combined to a Kalman filter, to estimate and correct IMU biases and inertial navigation system errors, in order to overcome fast drift problems of the low-cost INS, then improve navigation accuracy [11] [12]. …”