Sensor Fusion Based on Integrated Navigation Data of Sea Surface Vehicle with Machine Learning Method
Egemen Belge,Recep Fatih Cantekin,Berna Erol,Volkan Akgül,Seda Karadeniz Kartal,Rifat Hacioğlu,Sedar Kurtuluş Görmüş,Şenol Hakan Kutoğlu,Kemal Leblebicioğlu,Volkan Akgul,Rifat Hacioglu,Sedar Kurtulus Gormus,Senol Hakan Kutoglu,Kemal Leblebicioglu
DOI: https://doi.org/10.1109/inista52262.2021.9548442
2021-08-25
Abstract:Underwater mapping is important for many studies such as underwater cable/pipe platform placement and monitoring, bridge piers placement, dam construction, geological and geophysical studies. The position and orientation information of the sea surface vehicle is measured from the Global Positioning System (GPS) and Inertial Measurement Unit (IMU) sensors are placed on the vehicle, and the height from the seafloor is measured from a single beam sonar. External disturbance effects such as waves and wind cause oscillations in the surface vehicle. In bathymetric measurements, which are of great importance in mapping, errors due to oscillations occur. In underwater mapping, the orientation effect of the sea surface vehicle should be known in order to minimize these errors. In the study, to minimize these error sources, data from 3 different IMUs integrated into the sea surface vehicle are fused with sensor fusion algorithms such as the Integrated Navigation System (INS) and Support Vector Machine (SVM). In this study, a machine learning-based SVM integrated navigation system with minimum error in optimum positioning of the surface vehicle under external disturbance effects is proposed. With the INS, the performance of the machine learning-based SVM sensor fusion algorithm is analyzed comparatively.