Dead reckoning-aided GPS vehicle navigation using sequential least squares with weighted constraints

Nakarin Satthamnuwong, Purdue University

Abstract

This dissertation deals with data processing, kinematics models, simulations, and a prototype design for a low-cost positioning system using the Global Positioning System (GPS) and a dead reckoning system. The extended derivation from the conventional Least Squares leads to the sequential least squares for a static and kinematic model in non-linear problems. The final derivation has proven to be identical to the famous Kalman Filter mainly used in the electrical and control engineering area. The least squares method provides a better understanding of the kinematic model for surveyors who are familiar with the lease squares methods in general without gaining knowledge of random process and differential equations, which are the foundation of the Kalman filter. The GPS data reduction for C/A code pseudorange equations, sensors, and mathematic models for the dead reckoning system are discussed. The prototype system for the dead reckoning-aided GPS navigation in this research consists of off-the-shelf equipment, such as a laptop computer, a C/A code GPS receiver, distance sensors, and an electronic compass. Simulations in two-dimensional navigation show that the accuracy of the positioning depends on the precision of the observations, the level of integration in the kinematics model, and the actual motions of the receiver in respect to the model. Since the motion of the receiver is unpredictable for the model, a testing procedure for the predicted residuals is required to prevent large errors from the unmatched model. The field experiments for the prototype system have been performed at the Purdue University Airport. High-accuracy dual-frequency GPS receivers verify the accuracy of the prototype system.

Degree

Ph.D.

Advisors

Gelder, Purdue University.

Subject Area

Civil engineering

COinS