AUV positioning based on interactive multiple model

The research topic of autonomous underwater vehicles (AUVs) has attracted much attention over years since they provide marine researchers easy ways to access the ocean for surveying and site investigation, etc. To accomplish these applications, an AUV has to know its position accurately. Therefore, AUV localization is very important problem. In this paper, we propose an interactive multiple model (IMM)-based method for AUV localization because this method is capable of tackling complex behaviors of vehicles with different dynamic models. Several filtering techniques, namely, Kalman filter (KF), particle filter (PF) and modified PF (MPF), are investigated to estimate the position of the AUV. In development of the MPF, an ℓ1-norm is used to compute particle's cost instead of their weight to allow us to operate the filter without the use of measurement information. Those filters are running in parallel and the estimates are integrated by the IMM-based algorithm to obtain the position of the AUV. The sensor unit onboard consists of global positioning system (GPS), Doppler velocity log (DVL), inertial measurement unit (IMU) and a digital compass. Different dynamic models are studied to demonstrate the performance of the IMM-based methods, namely, IMM-KF, IMM-PF and IMM-MPF. Field trials using the STARFISH AUV show the capability of the algorithm.