Map aided SLAM in neighbourhood environments

Robust and accurate localization is a very important issue for the application of smart vehicles in neighbourhood environments such as theme parks, industrial estates, university campuses, etc. Conventional and classical approaches based on global positioning system (GPS) when used in closed spaces like neighbourhood environments pose problems due to signal blockages and multiple path effects. Feature based localization techniques suffer from feature detection failures, especially when features are sparse or not recognisable. Dead reckoning and inertial methods have to deal with the problem of drift in the sensors to be able to localize reliably over long periods of operation. To localize a vehicle reliably, robustly and accurately, a framework that enables the fusion of the different localization techniques is thus required, for this purpose, a road network topology constrained unified localization scheme is proposed based on the general Bayesian probabilistic estimation theoretic framework. The experimental results obtained from a vehicle driven in a large neighbourhood environment are presented to demonstrate the effectiveness of the proposed methodology.