Investigation of A New Method for Improving Individual Vehicle Speeds Estimation with Advanced Loop Data

In this study, a previous speed estimation model supported by advanced loop data was evaluated in a new dataset with a high volume of large vehicles. With 23% large vehicles in the traffic stream, the performance of the previous model got worse. Duration of a vehicle traversing the loop sensor, which was the feature selected to develop the previous model was easily influenced by the variation of vehicle lengths. Leading edge duration which indicates the time period from when the front bumper of the vehicle first gets onto the loop to the vehicle fully covering the loop was investigated. A new speed estimation model with a new feature (leading edge duration) was developed. The leading edge duration was justified as a more reliable feature than duration to be used in speed estimation. Two models based upon the corresponding features were compared. Finally, results of the model comparison and suggestions for the further model improvements were provided