Enhancing INS Accuracy in GNSS-Denied Environments: Incorporating Vehicle Dynamics Motion Models and Slip Estimation in INS Algorithm to Improve Positioning Accuracy

Abstract

Accurate vehicle positioning is critical for navigation systems, and one approach to gain high positioning accuracy is to fuse the measurements from a global navigation satellite system (GNSS) and an inertial measurement unit (IMU). However, certain challenges remain, particularly in scenarios involving vehicle slip and GNSS-denied environments. This thesis investigates how using vehicle-specific motion models and slip estimation can enhance the performance of an inertial navigation system (INS). Three motion models are developed and evaluated in both simulated and real-world scenarios. These are a constant acceleration model (CM), a unicycle model (UM), and a bicycle model (BM). A slip estimation method is proposed, using a Kalman filter to adaptively estimate slip parameters based on GNSS and IMU data. Results show that the UM and BM outperform the CM in most scenarios, with the BM demonstrating superior accuracy in the presence of slip. Real-world tests show potential for the UM and BM as they seem to be able to follow the expected shape of the trajectory. However, limitations due to 2D assumptions result in incorrect values for the velocity, leading to inadequate positioning accuracy. The results indicate that integrating advanced motion models and slip estimation into INS algorithms can significantly improve positioning accuracy. However, further testing and extension to 3D implementations are necessary to validate these results in real-world applications.