ABSTRACT
Vehicle pose estimation with high bandwidth and long-term accuracy usually involves data fusion of different sensors because there is no single sensor to satisfy both requirements. In this respect, global positioning system (GPS) or vision systems in conjunction with inertial measurement unit (IMU) are considered complementary positioning systems: GPS or vision systems provide low update rate, but they have the advantage of long-term position accuracy and stability. Conversely, IMU systems provide high bandwidth position information, while they are characterized by longterm drift. Efficient data fusion process requires the statistical characteristics of the measurement noises. However, the measurement error associated with a GPS devices often varies from one point to the next, so the error depends on many factors such as satellite geometry, atmospheric condition, multipath areas, and shadow. This means that the covariance matrix associated with the GPS noise is not known beforehand and, therefore, has to be estimated in the internal model for automatic tuning. This chapter focuses on the integration of IMU with two real-time kinematic (RTK) GPS units in an adaptive Kalman filter (KF) for driftless estimation of a vehicle’s attitude and position in three dimensions. Comprehensive nonlinear models of the dynamics and observation of the estimation system are presented, followed by application of a linearization technique. The discrete-time model, which involves a state-transition matrix and the covariance of process noise, is derived in closed form, thus rendering the filter suitable for real-time implementation. Additionally, integration of the
21.1 Introduction ......................................................................................................................... 367 21.2 Vehicle Pose Estimation: A Brief Review .......................................................................... 369 21.3 Modeling and Analysis ....................................................................................................... 369
21.3.1 Observation and Process Models .......................................................................... 370 21.3.2 Observability Analysis .......................................................................................... 372
21.3.2.1 Dual-GPS/IMU Integration ................................................................. 373 21.3.2.2 Single-GPS/IMU Integration ............................................................... 374
21.4 Estimator Design................................................................................................................. 375 21.4.1 Estimation of the GPS Noise Covariance ............................................................. 377 21.4.2 Initialization of KF ............................................................................................... 378
21.5 Experiment .......................................................................................................................... 379 21.5.1 Driftless Attitude Determination Using Three GPS Antennas
(Triangulation Method) ......................................................................................... 379 21.5.2 Test Results ...........................................................................................................380
