Traffic Pattern Analysis and Anomaly Detection Using Large-scale Trajectory Data

Monday, June 17, 2024

1:30 PM - 2:45 PM ET

Room: Galleria

Primary Author(s)

Desh Sharma

Graduate Student
Kennesaw State University
Marietta, GA, United States

Co-Author(s)

CT

Christian Taylor, n/a

Graduate Student
Kennesaw State University
Marietta, Georgia, United States
JP

Janki Patel, MS (she/her/hers)

Graduate Student
Kennesaw State University
Gray, GA, United States

Presenter(s)

Junxuan Zhao

Limited Term Assistant Professor
Kennesaw State University
Marietta, GA, United States

Co-Author(s)

JZ

Jack Zheng, n/a

Associate Professor
Kennesaw State University
Marietta, Georgia, United States

Submitter(s)

JZ

Junxuan Zhao (she/her/hers)

Limit Term Assistant Professor
Kennesaw State University, United States

Abstract:

With the rapid development of Internet-of-Thing (IoT) technology and the increasing computing capabilities, high-resolution high-accuracy real-time trajectories of motorized vehicles and vulnerable road users are now easier to obtain through advanced traffic sensing. Traditional approaches of manually checking trajectory datasets to extract valuable information are extremely time-consuming and error prone. In this study, we focus on using machine learning algorithms to analyze traffic patterns and detect anomalous driving events in extensive trajectory data. We aim to investigate effective and efficient techniques and procedures for processing time-series trajectory data and apply the proposed method to an open-source dataset (InD dataset) for a case study. The main procedure includes data preprocessing, feature extraction, machine learning model training and anomaly detection. Traffic patterns such as daily traffic trends, speed variation, and acceleration characteristics are analyzed at 4 intersections. The identified irregular paths demonstrate abnormal driving behavior including unexpected U-turns, unusual stops, and more. Our study not only reveals these anomalies but also explores their implications on traffic management and safety. Furthermore, we discuss potential practical applications in the realm of vehicle-to-infrastructure (V2I) alert systems.

Learning Objectives:

Attendees can expect to learn the following from this session:

Present large-scale trajectory data of different road users in an interactive visualization.
Analyze traffic patterns and detect abnormal driving events using machine learning algorithms.
Develop new potential applications in traffic management and warning/alert systems.