The framework combines visual information with data from an inertial measurement unit (IMU), using the robot’s self-dynamics to predict feature motion, represent coupled motion states, and correct ...