Design an‎d implementation of navigation system based on fusion of inertial, flight an‎d visual sensors

10/8/2025 12:00:00 AM

In this research, a new approach to navigation based on image data, inertial data, an‎d flight data is presented. This method is based on combining the image-inertial navigation method called Vins-Mono with a model extracted from the drone based on its stored flight data. The extracted model of the aircraft is constructed using a Multi-Layer Perceptron (MLP) neural network, an‎d it only estimates the linear position states, which deviate significantly over time. According to flight results with a hexacopter, whose model was calculated offline through multiple flights, it shows an improvement of approximately 42% in the Vins-Mono algorithm, meaning the error in position estimation is reduced from 4.29 meters to 2.46 meters.

فيلتر كالمن , ناوبري تصوير , اينرسي , داده پروازي , دوربين , ناوبري بدون GPS , شبكه هاي عصبي , ديناميك , پهپاد

Kalman filter , – image navigation , inertia , camera , GPS-free navigation , neural networks , dynamics , drone

Hamed Pourkaveh

Saeid Shamghadari