Design and Implementation of an Offline Map-Matching System for Floating Car and Telco Data

Nowadays, the presence of positioning devices enables the collection of detailed user and vehicle trajectories. However, due to the intrinsic inaccuracies of positioning systems, several preprocessing steps are required to correct trajectory errors. Among these, map matching plays a key role: it involves aligning raw positional data with the underlying road network to reconstruct the actual path taken. The goal of this thesis is to build a unified, adaptable map matching system that can operate effectively across datasets with varying characteristics by adapting algorithm parameters accordingly. Specifically, the focus is on the design and implementation of an offline map-matching algorithm, i.e., an approach that processes the entire trajectory only after all points have been collected. To achieve an efficient system for large-scale data, the final algorithm was implemented using PySpark DataFrames, enabling parallel data processing across multiple nodes or cores. The proposed algorithm was initially tested on Floating Car Data (FCD), a dataset characterized by high spatial and temporal resolution, then it was applied to a degraded version of the same data in order to simulate the behavior of Telco data and calibrate the algorithm parameters. Finally it was executed on the Telco dataset, which instead presents lower spatial accuracy and a wider sampling interval. The algorithm has been tested on a road network graph containing the main roads of Italy (i.e., motorways, primary roads, secondary roads, and trunk roads). The experimental results demonstrate that the algorithm achieves high accuracy on FCD data, particularly for longer trajectories, and produces plausible routes even under mildly degraded data. Even on Telco data, although its lower precision introduces additional uncertainty, the approach remains robust. Moreover, the execution time grows slowly with the number of processed trips, making the solution suitable for large-scale mobility analysis. This work provides a flexible and scalable framework that can be reused or extended for future applications in traffic monitoring, mobility studies, and transportation planning.