An Enhanced GNSS/MINS Integrated Navigation System With a Wheel Velocity Predictor Based on Vehicle Dynamics Model

The GNSS/INS integrated navigation system is widely used in intelligent vehicles. Due to the vulnerability of GNSS and the low accuracy of low-cost inertial devices, the performance of dead reckoning during GNSS outages is of great concern. The non-holonomic constraint (NHC) is a widely used method to improve dead reckoning performance, but it ignores the sideslip effect caused by tire deformation, which leads to performance degradation. The conventional vehicle dynamics model can achieve higher accuracy by modeling tire sideslip effects; however, its parameters are often complex and challenging to obtain. This work proposes a novel wheel velocity predictor (WVP) based on the vehicle dynamics model, which models tire sideslip effects while having concise and easy-to-estimate parameters. Based on this model, an integrated navigation system is proposed, which can adaptively estimate WVP parameters when GNSS is available and use the additional information from trained WVP to improve navigation performance during GNSS outages. Notably, the proposed system utilizes a more accurate model than conventional NHC while avoiding requiring additional sensors or complex prior vehicle dynamics model parameters. Field experiments have shown that the proposed scheme can accurately estimate the tire sideslip angle in real-time and significantly improve dead reckoning performance compared with the NHC-based method. Furthermore, as part of the WVP parameters, the inertial measurement unit mounting angles can be estimated even with a non-straight driving trajectory. Compared with the NHC-based method, which requires a straight driving trajectory to converge, the application limitation is significantly relaxed.

I. Introduction

In recent years, there has been a higher demand for accuracy, integrity, and continuity of position in many intelligent transportation applications such as autonomous driving [1], [2] and cooperative-intelligent transportation systems [3]. Due to its accuracy and global coverage, the Global Navigation Satellite System (GNSS) has become the fundamental means for vehicle positioning. Several emerging GNSS high-precision positioning techniques [4], [5], [6] can achieve centimeter-level precision and have been adopted by more and more high-level driving assistance systems. However, GNSS satellite signals are susceptible to fluctuations in electromagnetic propagation environments, particularly in complex road environments such as urban canyons, tunnels, viaducts, etc. In such cases, the GNSS may become unusable. As inertial navigation system (INS) and GNSS have complementary characteristics, the GNSS/INS integrated navigation system has become the most commonly used solution. [7]. During a GNSS signal outage, the INS dead reckoning (DR) accuracy will decline over time due to bias drifts and other errors of the inertial measurement unit (IMU). The worse is that, for mass-produced vehicles, low-cost micro-electro-mechanical system IMUs (MIMU), which have relatively large bias drift, are often the only option. Therefore, the main focus of such a system is dead reckoning performance during the GNSS outages. Several other sensors are introduced to improve the performance of such an integrated system based on low-cost MIMU during GNSS outage [8], [9], [10]. However, these methods directly increase the complexity and cost of the system.