A framework for on-demand creation of vector tiles for OpenStreetMap data

Abstract

Within the OpenStreetMap (OSM) project, the traditional approach for delivering map tiles relies on raster tiles. To ensure fast response times, a subset of the raster tiles, typically corresponding to low zoom levels is prerendered and stored on disk. However, this preparation step is time-consuming and demands significant storage resources. For instance, storing all tiles up to a zoom level of 14 requires approximately 200 GB of disk space [1]. Furthermore, as OSM data is frequently updated, this process must be repeated occasionally to ensure an up-to-date map. In contrast, vector tiles offer a modern alternative to raster tiles, as they are rendered on the client side and can be generated much more quickly. Instead of relying on existing vector tile services with fixed map generalization strategies, the goal of this work is the development of a framework for on-demand creation of vector tiles for OpenStreetMap Data, by leveraging query data structures like in [2]. To achieve this, we begin with a review of the relevant background and related work, covering key concepts in web mapping, including the Web Mercator projection and the Mapbox Vector Tile (MVT) format. We also introduce the shortest path problem and Contraction Hierarchies (CH), and briefly survey existing approaches. Based on this foundation, Chapter 3 presents the conceptual design of the framework, detailing its key components. In Chapter 4, we present a prototype implementation of the proposed framework. This prototype integrates the query data structure introduced in [2], which is described in more detail. Furthermore, we address key challenges and solutions related to the required spatial indexing structure, the tile creation process, and the selection of suitable LoD candidates. Chapter 5 provides an evaluation of the methods implemented in the prototype, followed by the conclusion and outlook.