Large Data Analysis: How Can Accidents Be Detected Based on Probe Vehicle Data?

Accidents continue to affect both the people involved and the flow of traffic. Timely and accurate accident detection can greatly benefit emergency services and traffic management, as a delay of minutes can determine the difference between life and death. This study introduces a method for detecting accidents using floating car data. The algorithm analyzes vehicle trajectories based on criteria derived from Kerner’s three-phase traffic theory, determining a high likelihood of an accident at a specific time and location. Validation using third-party data confirms the occurrence of real accidents. Empirical examples from a large feet of connected vehicles demonstrate the method’s effectiveness: it can swiftly and accurately detect freeway accidents, distinguishing them from normal congestion. A median improvement in detection time of 6.5 minutes is achieved compared to ground truth.