[論文レビュー] BigSUMO: A Scalable Framework for Big Data Traffic Analytics and Parallel Simulation

Rahul Sengupta, Nooshin Yousefzadeh+8 more|arXiv (Cornell University)|Jan 5, 2026

Traffic Prediction and Management Techniques被引用数 0

ひとこと要約

BigSUMO is an end-to-end, open-source framework for city-scale analytics and parallel SUMO-based simulations using ATSPM and sparse trajectory data, enabling descriptive analytics, interruption detection, and prescriptive traffic optimization.

ABSTRACT

With growing urbanization worldwide, efficient management of traffic infrastructure is critical for transportation agencies and city planners. It is essential to have tools that help analyze large volumes of stored traffic data and make effective interventions. To address this need, we present ``BigSUMO", an end-to-end, scalable, open-source framework for analytics, interruption detection, and parallel traffic simulation. Our system ingests high-resolution loop detector and signal state data, along with sparse probe trajectory data. It first performs descriptive analytics and detects potential interruptions. It then uses the SUMO microsimulator for prescriptive analytics, testing hundreds of what-if scenarios to optimize traffic performance. The modular design allows integration of different algorithms for data processing and outlier detection. Built using open-source software and libraries, the pipeline is cost-effective, scalable, and easy to deploy. We hope BigSUMO will be a valuable aid in developing smart city mobility solutions.

研究の動機と目的

• Provide an end-to-end, scalable analytics and simulation pipeline for ATSPM and sparse trajectory data.
• Deliver descriptive analytics via interactive notebooks and prescriptive analytics through parallel SUMO simulations.
• Enable interruption detection from multiple data sources and support counterfactual analysis of interventions.
• Utilize open-source components for cloud deployment and broad accessibility.
• Demonstrate application workflows and potential use cases in smart city mobility solutions.

提案手法

• Ingest ATSPM and sparse trajectory data and apply descriptive analytics to extract wait times, turning movements, queue lengths, and braking events.
• Generate spatial masks from GIS data to clip trajectories and preprocess data with GeoPandas and MovingPandas.
• Detect interruptions by combining ATSPM-based and trajectory-based outlier analyses using ABOD in PyOD, run on parallel intersections.
• Calibrate and run hundreds of SUMO microsimulations in parallel to evaluate counterfactual traffic interventions.
• Use a basemap and O-D/turn movements to calibrate flows and speeds for realistic SUMO simulations, leveraging routeSampler for traffic generation.
• Execute parallel simulations on commodity CPUs to produce large-scale data for model training and evaluation.

実験結果

リサーチクエスチョン

• RQ1 How can ATSPM and sparse trajectory data be integrated to produce city-scale descriptive and prescriptive traffic analytics?
• RQ2 What are effective methods to detect traffic interruptions using heterogeneous data sources in a scalable pipeline?
• RQ3 Can parallel SUMO microsimulations under varying parameters reliably support counterfactual analysis and intervention planning?
• RQ4 What preprocessing, masking, and calibration steps are necessary to align simulation with observed traffic behavior?
• RQ5 How can this framework support data generation for deep learning models in transportation?

主な発見

• The BigSUMO pipeline reaches 2–3 minutes wall-clock time per intersection for end-to-end processing including clipping, analytics, vectorization, and outlier detection.
• The trajectory-based interruption detection is feasible and complementary to ATSPM-based detection, with parallelization enabling multi-intersection analysis on commodity CPUs.
• SUMO-based prescriptive analytics can run hundreds of parallel simulations to explore counterfactual scenarios for corridor optimization.
• Calibrated basemaps, O-D matrices, and speed distributions enable realistic SUMO simulations and faster convergence to observed traffic patterns.
• The framework supports data generation for deep learning, including graph-based models for spatial-temporal traffic dynamics.