Benchmark comparativo de ferramentas de código aberto para map matching online

Abstract

The process of mapping GPS measurements, which are often inaccurate, noisy, or sparse, to the corresponding road segment in a digital map is known as map matching. With the significant growth in the number of mobile devices equipped with geolocation sensors, this procedure has become a cornerstone for numerous modern applications, being directly constrained by the accuracy and efficiency of the algorithmic approaches employed. Map matching methods are divided into two main paradigms: online, in which the route is processed incrementally, and offline, in which the complete trajectory is analyzed as a post-processing step. As a result of this relevance, a broad ecosystem of open-source tools has emerged, offering multiple alternatives to developers and researchers. However, the informed selection of an appropriate solution, especially for online scenarios, is hindered by the lack of comparative, reproducible benchmarks and by the absence of systematic empirical evaluations of tools originally offline when adapted for incremental execution. In this context, this work develops and applies a systematic and reproducible benchmark for evaluating open-source online map matching tools, including natively offline solutions adapted through techniques such as mini-batch processing. As main contributions, we present an automated benchmark harness with container-based orchestration, the mmlib library, which standardizes the client interface and enables incremental execution of the tools, and a synthetic dataset with controlled ground truth, generated via urban traffic simulation. The evaluated tools are compared in a multidimensional manner, considering spatial accuracy, latency, computational resource consumption, and continuous processing capability. The obtained results provide empirical evidence that helps software engineers make informed choices of solutions for real-time applications and offer an extensible basis for evaluating new approaches and tools in the map matching domain.