|
M.S. Defense Thursday, December 14, 2006 Eng II - Room 205 4:00 pm Title: Time Series Analysis of Vehicle Velocity Data for the Identification of Freeway Level of Service Abstract: In many parts of the country, Freeway Performance Measurement Systems (PeMS) are being implemented to collect real time data from our congested freeways. For example, California has an elaborate PeMS that uses thirty-second aggregated time and traffic data recorded from embedded loop detectors in the freeway of interest and averages the data from thirty-second intervals to five-minute aggregates. The loop detectors are collecting statistics for traffic speed, density, and occupancy, the average length of time a vehicle is passing over the loop. With these statistics, PeMS calculates other information such as flow rate, g-factor, vehicle miles traveled, vehicle hours traveled, and level of service (LOS). The California PeMS website provides easy access to charts and graphs of calculations and can even output the LOS for a given day or range of days and times on a specific area of the freeway. The freeway Level of Service (LOS) is an indicator of traffic flow and the health of the system. The PeMS system collects macroscopic data and using that data calculates and relates all of the data to each other. At the microscopic level, it is of interest to see how independently collected data (such as velocity trajectories from a single vehicle) relate to the macroscopic level. This thesis develops a method to relate microscopic and macroscopic traffic data by developing a pattern recognition system capable of recognizing and directly relating vehicle speed trajectories to the Level of Service. The approach described in this thesis uses microscopic vehicle speed vs. time data, (i.e. second by second) as input into a time series analysis system to determine Level of Service. This approach is based on a recognition system that requires a predetermined training set of vehicle speed data with known Level of Service categories A-F. The system is built on the use of wavelet transforms to extract useful information from the vehicle speed data. Experiments show a upwards of a 90% success rate in pattern matching local microscopic speed trajectories to macroscopic Level of Service. |
| < Prev | Next > |
|---|