Work

Optimising low-cost GNSS positioning for road constrained vehicles using a Bayesian filter

Year: 2018
Type: dissertation
Abstract: Co-operative Intelligent Transportation Systems (C-ITS) are a growing industry that have significant implications for the safety and efficiency of on-road vehicle use.Accurate and reliable vehicle positioning is a critical component in C-ITS in order for the component vehicles and infrastructure to interact effectively and information-share for collectively enhanced services.Such services include vehicle navigation, collision-avoidance systems, driver assistance and emergency notifications (for example, mobilised if an ambulance needs to pass through the street occupied by the positioned vehicle).A range of existing and emerging technologies have been developed and tailored to improve the accuracy and continuity of positioning for C-ITS, which include Global Navigation Satellite Systems (GNSS), digital road maps, inertial sensor technologies, dedicated short-range communication (DSRC) and other technologies.However many of these technologies are expensive and in many cases prohibitive for large scale rollout and general public use.Therefore it is important to develop and optimise low-cost vehicle positioning methods to complement more expensive technologies in order to assist with the large scale feasibility of C-ITS, as well as meet individual-scale immediate user needs at viable cost.The subject of this PhD research is optimising low-cost vehicle positioning; that is maximising the accuracy and continuity of position estimates generated from exclusively low-cost data sources.The focus of this research is improving mathematical methods for data fusion (the statistical combination of diverse data inputs for estimation), and a new Bayesian approach is presented for real-time vehicle positioning from the data fusion of low-cost GNSS data and freely available digital road map data.The design of this new method, the Gaussian Cluster Approximate Filter (GCAF), is driven in part by a detailed study of common shortcomings and the causes of failure-modes of existing methods in a range of common urban vehicular situations, in order to be robust to these positioning challenges.Through robust testing it is demonstrated that effective combination of these two low-cost data sources using statistically robust mathematical methods provides substantial improvement to existing approaches, and lessens the need for more expensive sensors in many common vehicular environments.The method is designed to be easily modified to allow fusion with other data sources as future applications may require.
Language: English
Open Access status: green
OpenAlex ID: W2904707528
Cites: 52
Cited by:
Related to: 10