Map-based localization of mobile vehicles

Abstract

As mobile vehicles are getting more and more autonomous, the positioning capability becomes an important issue because it is the key technology of navigation. There are two kinds of localization methods. One is a dead eckoning and the other is an absolute positioning. These two methods have their own advantages as well as disadvantages. So the fusion of two methods is desired. In this thesis, map-based localization algorithms for an indoor and an outdoor mobile vehicles are discussed. To estimate the robot pose using a map, the map structure should be decided. The generalized symmetric perturbation map (GSP map) using a line pose is proposed. And the error characteristic of the line feature is studied based on the error of the laser scanner. In the map-based localization, there are three different situations. The first situation is when the prior pose of the robot is unknown and the second situation is when the prior pose is known and the third situation is when the global map is incorrect. We propose estimation algorithms for each situation using the GSP map. We propose a distributed estimator for cooperative localization. We show the effectiveness by several simulations and experiments. For an outdoor mobile vehicle such as a car, GPS is most widely used. But it has a large error if not corrected. To correct this error, a map matching method is introduced, which uses a road map to correct the position error. In this thesis, a novel adaptive fuzzy network (AFN) based C-measure algorithm is proposed, which can find the position of a car in the road map precisely. The C-measure algorithm is easy to calculate, and the calculation time does not increase exponentially with the increase in the number of junctions. For experimental tests, a car navigation system is built with a DSP chip, a GPS receiver, a vehicle speed sensor and a gyroscope. The real road experiments validate the effectiveness and applicability of the proposed algorithm and the developed car navig...