Road vehicle state estimation using low-cost GPS/INS

Abstract

Due to noise and bias in the Inertial Navigation System (INS), vehicle dynamics measurements using the INS are inaccurate. Although alternative methods involving the integration of INS with accurate Global Positioning System (GPS) exist and are accurate, this kind of system is far too expensive to become value-adding to production vehicles. This thesis therefore considers two aspects: 1) the possibility of estimating vehicle dynamics using low-cost INS and GPS, and 2) the importance of vehicle dynamics in terms of handling in the eyes of customers upon vehicle purchase. The former aspect is considered from an engineering perspective and the latter is studied in a marketing context. From an engineering point of view, knowledge of vehicle dynamics not only improves existing safety control systems, such the Anti-lock Braking System (ABS) and Electronic Stabilising Program (ESP), but also allows the development of new systems. Based on modelling and simulation in MATLAB/Simulink, low-cost GPS and in-car INS (such as accelerometers, gyroscopes and wheel speed sensors) measurements are fused using Kalman Filters (KFs) to estimate the vehicle dynamics. These estimations are then compared with the simulation results from IPG Car- Maker. For most simulations, the speed of the vehicle is kept between 15 to 55kph. It is found that while triple KF designs are able to estimate the tyre radius, the longitudinal velocity and the heading angle accurately, an integrated KF design with known vehicle parameters is also able to estimate the lateral velocity precisely. Apart from studying and comparing different KF designs with restricted sensors quality, the effects and benefits of different sensor qualities in dynamic estimations are also studied via the variation of sensor sampling rates and accuracies. This investigation produces a design procedure and estimation error analyses (theoretical and graphical) which may help future engineers in designing their KFs. From a marketing perspective, it is important to understand customers’ purchase reasons in order to allocate resources more efficiently and effectively. As GPS/INS KF designs are able to enhance vehicle handling, it is vital to understand the relative importance of vehicle handling as a consumer purchase choice criterion. Based on two surveys, namely the New Vehicle Experience Survey in the US (NVES US) and the New Car Buyer Survey in the UK (NCBS UK), analyses are performed in a computer program called the Predictive Analytics SoftWare (PASW), which is formerly known as the Statistical Package for the Social Sciences (SPSS). The number of purchase reasons are first reduced with factor analysis, the latent factors produced are then used in the SPSS Two Step Cluster analysis for customer segmentation. With the customer segments and the latent factors defined, a discriminant analysis is carried out to determine customer type in the automobile sector, in particular for Jaguar Cars. It is found that customers in general take vehicle handling for granted and often underrate its importance in their purchase. New vehicle handling-aided systems therefore need to be marketed in terms of the value they add to other benefits such as reliability and performance in order to increase sales and stakeholder value.