Drone navigation and localization. (IMU-based + vision-based)

First, we need an IMU based odometry. Basically, we need a nice dynamic model of a drone and IMU then applying Kalman Filter The expectation is to see big error by the time goes and when you perform some maneuvers with the drone. Then we will concentrate on basic visual-odometer that we need to evaluate the instantaneous velocities (at two or three dimensions) of the vehicle through optical flows. you will need to deal with very basic image processing. Then, let's apply fusion algorithms with the IMU-based odometer (by comparing tightly coupled and loosely coupled architectures). The expectation will be the enhancement in localizing our vehicle(drone). We should estimate the drone’s pose and build a consistent three-dimensional map. The results should be clear and report must be written preferably in dissertation-like style.