Studies on GPS INS integration using model based and soft computing approaches

Abstract

Global Positioning System GPS is a widely used satellite navigation system However GPS will not provide a continuous and reliable positioning at all the times as it is likely to be observed by buildings mountains etc Inertial Navigation System INS provides continuous information of position velocity and altitude at all the times However the performance of INS deteriorates with time due to the performance about the inertial sensors GPS INS Integration provides a reliable navigation solution by overcoming each of these shortcomings when it acts as a stand alone system including signal blockage in case of GPS and increased positional errors with time for INS Though GPS INS Integration has been attempted by several researchers it still faces challenges especially where navigation has to be done for maneuvering targets Existing GPS INS Integration using Kalman Filter KF can give correct results only when the system dynamic models are completely known newlineFor highly maneuvering targets navigation is provided using Interactive Multiple Model IMM The existing IMM Unscented Kalman Filter UKF with Constant Acceleration CA model and Coordinated Turn CT model is connected in parallel with an appropriate switching probability IMM obtains its estimate as a weighted sum of individual estimates from the 4 filters matched to different motion models of the target In order to further enhance the performance of IMM UKF this thesis proposes a Two Filter Smoothing TFS based IMM UKF The proposed IMM UKF TFS uses forward and backward smoothing to further improve the positional accuracy by reducing navigation error newline newline