Nonlinear filtering for tightly coupled RISS/GPS integration

Abstract

The integration of Global Positioning System (GPS), inertial sensors and other motion sensors inside land vehicles enable reliable positioning in challenging GPS environments. GPS signals may suffer from blockage in urban canyons and tunnels resulting in interrupted positioning information. Inertial sensors are standalone sensors that can be integrated with GPS and can bridge the blockage periods as they do not rely on any external signals. Recently, miniaturized Micro-Electro-Mechanical Systems (MEMS)-based inertial sensors are abundantly used for vehicle safety applications such as air-bag deployment, roll-over detection, etc. These sensors can be used as inertial navigation system (INS) after integrating with GPS for reliable navigation solution even in denied GPS signal environments. The traditional technique for this integration is based on Kalman filter (KF) with a dedicated inertial sensor module consisting of three orthogonal gyros and three orthogonal accelerometers. This research targets a low cost navigation solution for land vehicles and hence it utilizes a reduced inertial sensor system (RISS) consisting of MEMS-based single axis gyro and a dual axis accelerometer. Additionally, the vehicle's odometer is used and an integrated 3D navigation solution is achieved. To improve the positioning accuracy a nonlinear filtering technique, particle filter (PF) is used to avoid linearization errors. Because of PF ability to deal directly with nonlinear models, it can accommodate arbitrary sensor characteristics and motion dynamics. Consequently, tightly coupled integration which has a nonlinear measurement model can be directly used in PF without introducing any errors. An enhanced version of PF is implemented known as Mixture PF and the performance of this method is examined by actual road tests in a land vehicle and compared with KF.