Low-cost GNSS/INS/Odometric Sensor Fusion Platform for Ground Intelligent Transportation Systems

Abstract

For many applications in Intelligent Transport Systems (ITS), the position and heading information of vehicles and Vulnerable Road Users (VRU) cannot generally rely on the performance of the Global Navigation Satellite System (GNSS) as a standalone technology. Even with the use of precise positioning techniques and augmentation systems for GNSS, such as Space-based Augmentation System (SBAS) or Differential-GNSS, its performance and its availability still depends on the signal propagation conditions (e.g. multipath, unintentional or intentional (jamming) interferences, or visibility of satellites). The urban canyon represents one of the most challenging scenarios for GNSS standalone positioning, being a scenario where ITS users usually require the highest performance. This paper discusses the design, implementation and performance validation of multisensor positioning based on GNSS, Inertial Measurement Units (IMU), and Odometric information for ground ITS applications. Bayesian sensor fusion algorithms are discussed and loose and tight GNSS/INS coupling compared. Additionally, these algorithms are enhanced by exploring the application of dynamic noise covariance matrices, including non-holonomic constraints to the vehicle’s movement, and by adding a zero velocity update IMU calibration algorithm using the vehicle’s speedometer measurements. The algorithms performance was extensively evaluated in both simulation and in real-life experiments. The algorithms are validated in a low-cost prototype implementation. The prototype receiver, operating in real-time, is based on the popular Raspberry PI3 platform, a dual-IMU MEMS peripheral and a consumer-grade GNSS receiver. The paper includes a discussion of the implementation trade-offs, challenges, and adopted solutions. A measurement campaign where the developed prototype was mounted on a vehicle is discussed, showing the potential of this approach.