Irregularly Sampled Transit Vehicles Used as a Probe Vehicle Traffic Sensor

摘要

Performance monitoring is an issue of growing concern both nationally and in211u001eWashington State. Travel-times and speeds are a key measure in performance. In 211u001ethis project, the authors use vehicles as probes and develop a framework for 211u001emodeling the time series that arise from sampling transit vehicle locations as a 211u001efunction of time. These samples of vehicles' location are obtained from the King 211u001eCounty Metro Automatic Vehicle Location (AVL) system. An optimal filter method is 211u001edeveloped that estimates speed as a function of space and time. In this work, an 211u001eoptimal solution for the state vector, containing the variables' speed and 211u001eposition, is possible at each step using the Kalman filter result. This type of 211u001efilter solution requires the creation of a model for the process; in this case, a 211u001erelationship between location and time for the vehicles and the creation of a 211u001emeasurement model to account for measurement errors. The use of such formalism 211u001edepends upon the assumption that the deviations of the actual system from the 211u001eidealized model are normally distributed. The model was applied against data from 211u001eboth freeways and arterials to test this assumption. In most ranges of travel, 211u001ethe resulting probability of distribution membership is on the order of 0.9, 211u001eindicating that the assumption of normally distributed errors is indeed a good 211u001eone.