HVIOnet: A deep learning based hybrid visual-inertial odometry approach for unmanned aerial system position estimation

Muhammet Fatih Aslan, Akif Durdu, Abdullah Yusefi, Alper Yilmaz

Abstract: Highlights•A hybrid architecture consisting of CNN and multiple BiLSTM layers is proposed.•Efficient visual–inertial data fusion based on CNN-BiLSTM.•Proving the effectiveness of the method with simulation data in the ROS environment.•This algorithm performs efficient and easy UAS position estimation.