Model-aided and vision-based navigation for an aerial robot in real-time application

In this paper, a novel navigation method with the assistance of a vehicle dynamic model (VDM), known as the VDM-aided navigation method, is introduced. This method is specifically designed for a subset of fixed-wing aerial robots within the broader category of unmanned aerial vehicles. Vision-based navigation (VBN) is employed to increase accuracy while maintaining reliability in Global Navigation Satellite System (GNSS) outages. In addition, an unscented Kalman filter (UKF) is used to estimate navigation parameters, including speed, position and attitude. This method uses the dynamic system as a process model and employs VBN, barometric altitude and vertical gyro as measurement inputs. In VBN, the method of scale-invariant feature transform is used as a method for image matching. To ensure the real-time capability of this method with the existing microprocessor, a hardware-in-the-loop (HIL) laboratory has been utilized. According to nonlinear observability methods, one can show the proposed integrated nonlinear navigation is observable under all conditions. Finally, the results of the HIL laboratory demonstrate that the proposed approach can estimate the robot navigation parameters with an acceptable level of precision even in the absence of an Inertial Navigation System (INS) and GNSS. It was validated even when there was an error of up to 20% in VDM parameters. Furthermore, an investigation was carried out regarding the use of Extended Kalman Filter instead of the UKF for the integrated navigation output. In GNSS outage conditions, considering both accuracy and cost, this method can serve as a valuable alternative for aerial robots. In addition, this approach can be recommended for INS fault detection with or without GNSS. Additionally, the integrated navigation provided can substitute the GNSS/INS system during fault conditions.