Detector Location Optimization On Transportation Network

Traffic detectors are a critical component for acquisition of real-time trafficdata. It is necessary to deploy traffic detectors in a selective way. On the other hand,in reality, it is ubiquitous and inevitable that detectors are subject to failure. Thispaper firstly considered traffic detector placement problem for urban transportationnetwork by taking detector failure into consideration. Secondly, detector locationproblem was investigated for freeway corridor with the purpose of minimizing traveltime estimation error. Detector failure was also considered in this case. Finally, amulti-objective detector location model was created focusing on four different trafficapplications. The major contents of this dissertation are detailed as follows:(1)Under the condition of stable traffic flow, with the purpose of maximizingOD pairs coverage and traffic flow coverage as many as possible, five differentaspects were taken into account:(a)Minimizing the number of traffic detectors withthe requirement of covering all OD pairs;(b)Designing detector placement pattern toachieve a high reliability.(c)Detector location problem considering existing detectors.(d)When the OD coverage is satisfied, try to cover as much traffic flow as possible.(e)Detector location problem considering detector failure and more traffic flowcoverage. For the first four questions, related integer programs were created, theresults obtained were analyzed. For the fifth problem, a two stage stochastic integerprogram model was created. Furthermore, to avoid extreme bad situation, a two stagestochastic integer program model based on CVaR component was proposed.Numerical experiment investigated the relationship between covered OD pairs, trafficflow and allowable traffic detectors. The location patterns for different number ofdetectors are also discussed. Sensitivity analysis was conducted. A two stage geneticalgorithm was proposed to solve this problem.(2)Using unidirectional freeway corridor as background, the objective is tominimize travel time estimation error via loop detector. A quadratic integer programmodel was developed. The model was reformulated into a liner integer programmodel via applying a linearization method. On the other hand, detector failure wasalso considered. Two stage stochastic models with or without CVaR component wereproposed. A sampling based genetic algorithm was developed for solving this problem. This approach paved the way for problems with large allowable number ofdetectors.(3)AVI readers were used in our model. A multi-objective detector locationoptimization was proposed. This problem was solved by applying a distance-basedmulti-objective evolution algorithm. A results comparison between single objectivemodel and multi-objective model was done. The resulted Pareto frontier was alsoinvestigated.