Navigation in Wheeled Mobile Robots Using Kalman Filter Augmented with Parallel Cascade Identification to Model Azimuth Error

Loading...
Thumbnail Image
Date
2013-06-13
Authors
Rahman, Atif
Keyword
Navigation , Kalman Filter , Parallel Cascade Identificaton , Global Positioning System , Inertial Navigation System
Abstract
Unmanned ground mobile robots are land-based robots which do not have a human passenger on board. They can be either autonomous, or controlled via telecommunication. For navigational purposes, GPS is often used. However, the GPS signal can be distorted in obstructive environments such as tunnels, urban canyons, and dense forests. IMUs can be used to provide an internal navigational solution, free from external input. However, low cost IMUs are prone to various intrinsic sources of error, which leads to large errors in the long run. Using the short term accuracy of the IMU, and the long term accuracy of the GPS, these two technologies are often integrated to combine the aforementioned aspects of the two systems. For integration of the two, various methods are implemented. Such integration methods include Particle Filters, and Kalman Filters. Kalman Filters are commonly used due to their simplicity in calculations. However, the Kalman Filter linearizes the nonlinear error estimates which are inherent with low cost IMUs. The Kalman Filter also does not account for IMU measurement drift, which is present when the measurement unit is used for a long period of time. In this thesis, a Parallel Cascade Identification (PCI) algorithm is augmented with the Kalman Filter (KF) to model the nonlinear errors which are intrinsic to the low cost IMU. The method of integration used was 2D GPS/RISS loosely coupled integration using a Kalman Filter. The PCI algorithm modelled the nonlinear error for the z-axis gyroscope while the GPS signal was available. During a GPS outage, the PCI nonlinear error model was combined with the KF estimated error and the mechanization error, to provide a corrected azimuth. The KFPCI algorithm showed an improvement over the KF algorithm in RMS position error, maximum position error, RMS azimuth error, and maximum azimuth error by an average of 30.76%, 34.71%, 66.76%, and 53.58% in each of the respective areas.
External DOI