Pseudo DVL reconstruction by an evolutionary TS-fuzzy algorithm for ocean vehicles

Abstract

By development of ocean exploration, autonomous vehicles are employed to perform on-water and underwater tasks. Using an extended Kalman filter, Inertial Navigation System/Doppler Velocity Log (INS/DVL) integrated systems are trying to navigate in oceans and underwater environments when Global Positioning System (GPS) signals are not accessible. The dependency of DVL signals on acoustic environments may cause any DVL malfunction due to sea creatures or strong wave-absorbing material. In this paper, an improved version of evolutionary TS-fuzzy (eTS) is proposed in order to predict DVL sensor outputs at DVL malfunction moment, by utilizing an artificial intelligent (AI) aided integrated system. According to lack of input selection and shrinking, while the classic eTS suffers from soaring prediction errors and may result in instability, by adding these properties to eTS, the performance increases in long-term DVL outage. The proposed eTS-aided system makes ocean navigation purposes possible during long-term and simultaneous outage of GPS and DVL. These evolutionary fuzzy systems change their structure depending on the path which makes the trained fuzzy system more flexible with non-stationary and varying environments. The real sensor data is collected online with a test setup on a lake and then the algorithms are applied. The powerful capacity of the proposed data fusion method is demonstrated in analysis results.