GNSS/IMU Simulator for Position Estimation Using LOS-Based GNSS Modeling and Dynamic Allan Variance Gyro Simulation

In the current field of autonomous driving, a significant focus is on the development and utilization of simulators. However, existing autonomous driving simulators often lack thorough consideration of position estimation, particularly with regard to global navigation satellite system (GNSS) and inertial measurement unit (IMU). This study targets ground-moving robots and autonomous vehicles, aiming to develop a GNSS/IMU simulator capable of generating realistic errors that align with real-world environments to facilitate pre-validation of position estimation accuracy. Initially, for GNSS, positioning solutions were predicted using the number of satellites obtained through a line-of-sight/non-line-of-sight simulation based on 3D building data. Subsequently, to consider the performance degradation of the Allan deviation under dynamic environments, we improved the conventional gyro model based on static-environment Allan deviation by incorporating error factors associated with dynamic environmental influences. Finally, the accuracy of position estimation was verified using real vehicle and simulator-acquired sensor data. Similar trends in position estimation results confirmed the potential for pre-validation of position estimation accuracy.