This paper deals with a problem of map-based localization. When a robot has a pre-constructed map of an environment and explores there with sensing, the localization problem boils down to the problem of finding the closest data pattern to the sensing data from the map. Although, the map quality as well as the sensing data quality greatly affects to the localization performance, however it has been rarely discussed up to this point. To study the influence of the map quality to the localization performance, we propose a simulational experiment using an architectural 3D CAD models. We prepared a precisive 3D CAD model imitating a corridor of elementary school and generated a 3D wireframe map and virtual sensing images from the CAD model. The precisive 3D CAD model is regarded as the ground-truth data through the whole experiments. Then, we prepared several levels of broken maps with randomly shifted joinery parts such as, doors and windows. For localization process, our previously developed image-retrieval based localization method is used. Totally 3,500 number of localization experiments are conducted with various broken map setting. We show the influence of map-environment disparity to the localization performance by comparing these experimental results and discuss the requirements of map quality.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.