Topological Data Analysis for Urban Traffic Safety and Optimization Using Vehicle Trajectory Data

Topological Data Analysis for Advanced Urban Traffic Safety and Optimization Using Vehicle Trajectory Data 

Debbie Aisiana Indah (Corresponding Author)1,  Dr. Judith Mwakalonge2, Gurcan Comert3  Dr. Saidi Siuhi4. 

Abstract 

Urban transportation networks are dynamic ecosystems that continuously generate voluminous and high-dimensional vehicle trajectory data. Despite the need for precise and actionable insights from this data, traditional analytical methods often grapple with the intrinsic challenges posed by high-dimensional datasets, and their susceptibility to noise and outliers often impedes the accuracy and reliability of extracted information, limiting their practical applicability for predictive analysis and real-time decision-making. This paper presents a novel framework that integrates Topological Data Analysis (TDA) with machine learning models to extract and analyse essential topological features embedded within complex transportation datasets. TDA is a recent and fast-growing field, capable of capturing multi-scale insights from vehicle trajectory data that conventional data analysis methods may overlook. And unlike traditional techniques TDA exhibits remarkable resilience against noise and elegantly manages intricate correlations in high-dimensional spaces, unveiling patterns and structures that are necessary for predictive analysis. Our proposed framework aims to harness the robust capabilities of TDA with machine learning models to identify, capture and predict temporal patterns and anomalies that are indicative of near-collisions or potential disruptions within urban traffic flow. The detection of such patterns provides urban planners, policymakers, and transportation agencies with a toolset for early intervention, allowing for optimized traffic management and a significant reduction in accident risks. Furthermore, this analytical prowess holds significant promise for Connected and Autonomous Vehicles (CAVs). By incorporating the nuanced patterns detected through TDA into the decision-making processes of CAVs, they can achieve enhanced situational awareness, make safer navigation decisions, and adapt proactively to evolving traffic scenarios. By leveraging the robust capabilities of TDA for machine learning, our approach will help to ensure enhanced safety and operational efficiency within urban transportation networks. 

Keywords: Topological Data Analysis (TDA), Vehicle Trajectory Data, Machine Learning, Persistent Homology