Implementing Model Predictive Control based Variable Speed Limit on Urban Freeways: Data Imputation, Model Modification and Field Test Analysis

Abstract

Among different freeway traffic control strategies, Variable Speed Limit (VSL) shows its excellence in terms of control scale, technical feasibility and the capability of improving driving environment and traffic throughput. The Model Predictive Control (MPC) based VSL method provides a close form control loop enabling optimized variable speed limit value. The MPC-VSL control system relies heavily on a stable real time data source, an accurate traffic state prediction model and timely feedback from field implementation. The Vehicle Detection Stations (VDS) system is responsible for providing real time traffic flow related data. Most of the time VDS system works well, however, there are occasions when one set of loop lost data due to hardware failure, and this thesis provides imputation algorithm for missing data. The macroscopic traffic state prediction model in MPC-VSL control scheme is the modified METANET model. The feasibility of modifying one critical term in the original METANET model, namely “desire speed”, is tested in this thesis with different weather conditions using real field weather and loop detector data. The last part of thesis will be evolutionary analysis of VSL field test that was conducted on Whitemud Drive, Edmonton from August 13 to September 4 of 2015, borrowing the concept of time domain analysis scheme and system robustness analytical tool.