Performance Evaluation of Short Time Dead Reckoning for Navigation of an Autonomous Vehicle

Abstract: Utilizing a Global Navigation Satellite System (GNSS) together with an Inertial Navigation System (INS) is today a common integration method to obtain a positioning solution for autonomous systems. Both GNSS and INS have benefits and weaknesses where the best parts from both systems can be combined with a Kalman filter. Because of this complementary nature, it is of interest to look at the robustness of the positioning solution when the Global Navigation Satellite System is temporarily not available. The aim of the thesis has been to investigate different vehicle models and to evaluate their short-time performance using a Dead Reckoning approach. The goal has been to develop a system for a Heavy Duty Vehicle (HDV) and to find out for which time interval a specific model can stay within a certain range when the GNSS is lost. A GNSS outage could for example happen when driving on a highway and passing signs, bridges and especially when driving inside tunnels. Also, for a solution to become commercially interesting, it must be cheap. Therefore, is it common to use so called Micro-Electro-Mechanical-Systems (MEMS) sensors which are of low-cost but suffer from biases, scale factors and temperature dependencies which must be compensated for. The results from the tests show that some models are able to estimate the position with good precision during short time GNSS outages whereas other models do not deliver the required accuracy. The main conclusion is that care should be taken when choosing the vehicle model so that it fits its usage area and the complexity needed to describe its motion. There are also lots of parameters to look at when investigating the best solution, where modeling of the low-cost sensors is one of them.