We propose a low-cost integrated navigation system to enhance navigation accuracy using multiple sensors such as the inertial measurement unit (IMU) and Global Positioning System (GPS). Due to the heterogeneity of the employed sensors, the suggested navigation system is equipped with a graph-based optimizer formulated as a maximum a posteriori estimator. As already known, graphic optimization methods are computationally more complex. To reduce this complexity, high-frequency IMU measurements are pre-integrated between lower-frequency GPS measurements. These pre-integrated measurements can be utilized in conjunction with the magnetometer and barometer sensor measurements to carry out navigation in case of GPS weak signal reception. The resulting optimization graph will be solved using the marginalization technique to attain a less computationally intensive optimization favorable in real-time applications. Incorporating internal sensors in navigation will come with problems like the initialization phase. This problem is tackled by integrating the inertial measurements between the required states in the body frame of the last state and then transforming them into the world frame. To evaluate the applicability of the proposed method in real-time situations, a dataset collected by a SkywalkerX8 Unmanned Aerial Vehicle is utilized. This dataset includes all the necessary maneuvers, including climb, descent, turn and cruise during the fight time. The results confrm that our proposed integrated system has superior statistical measures and temporal performance compared to the extended Kalman flter (EKF); it outperforms the conventional Factor Graph Optimization and EKF by approximately 45%
We propose a low-cost integrated navigation system to enhance navigation accuracy using multiple sensors such as the inertial measurement unit (IMU) and Global Positioning System (GPS). Due to the heterogeneity of the employed sensors, the suggested navigation system is equipped with a graph-based optimizer formulated as a maximum a posteriori estimator. As already known, graphic optimization methods are computationally more complex. To reduce this complexity, high-frequency IMU measurements are pre-integrated between lower-frequency GPS measurements. These pre-integrated measurements can be utilized in conjunction with the magnetometer and barometer sensor measurements to carry out navigation in case of GPS weak signal reception. The resulting optimization graph will be solved using the marginalization technique to attain a less computationally intensive optimization favorable in real-time applications. Incorporating internal sensors in navigation will come with problems like the initialization phase. This problem is tackled by integrating the inertial measurements between the required states in the body frame of the last state and then transforming them into the world frame. To evaluate the applicability of the proposed method in real-time situations, a dataset collected by a SkywalkerX8 Unmanned Aerial Vehicle is utilized. This dataset includes all the necessary maneuvers, including climb, descent, turn and cruise during the fight time. The results confrm that our proposed integrated system has superior statistical measures and temporal performance compared to the extended Kalman flter (EKF); it outperforms the conventional Factor Graph Optimization and EKF by approximately 45%