Detection of Freeway Incidents Based on Vehicle Acceleration Measurements Using Connected Vehicle Data

Prompt incident detection is a key parameter in the incident management process. In this paper two automated freeway incident detection methods are proposed and tested using connected vehicle data. Vehicles acceleration behaviors in the front of the incident location were selected as a signature for incident detection. The first method utilizes the distribution of average acceleration in each segment under the no incident conditions and detects the abnormality in the acceleration behavior by comparing the measured test segment average acceleration with a percentile threshold of the used distribution. The second method presents an application of Likelihood Ratio Test (LRT) to freeway incident detection. This method has shown superior performance in detecting signal in a noisy environment in electrical engineering. The LRT method examines the likelihood of the measurements belonging to the incident case and no incident case. If the likelihood ratio is more than a threshold, the incident is detected in the corresponding segment. A microsimulation tool was used for testing the performance of the two methods with different parameters. The results show that the second method outperforms the first method, particularly under traffic breakdown conditions.