Applying Operations Research Techniques to Improve Motor Vehicle Crash Emergency Response and Traffic Monitoring Using Intelligent Transportation Sensors

Intelligent transportation sensors are widely used in transportation systems to improve emergency response system and provide real-time traffic and weather related information to road users and traffic managers. These ITS sensors are increasingly equipped with wireless networking capabilities that help in data assimilation and apply data fusion techniques. The goal of this dissertation is to improve the current state of the transportation system by applying operations research techniques to control intelligent transportation sensors.;In the first part of this dissertation, we consider the use of stationary and mobile ITS sensors from different classes in motor vehicle crash detection and characterization. For the class of stationary sensors, placement plays an important role in crash detection, due to the constraints on sensor detection radius. Operations research techniques formulate this placement of sensors as a quadratic maximal coverage location problem. To solve this formulation we use an explicit-implicit-simulation based hybrid heuristic designed for the sensor placement problem. The mathematical model approximates the real-road sensor behavior. To understand the real-world behavior of sensor placement we developed a simulation model using real-road data available from 2004-2009. Simulation results prove that the solution generated using a hybrid explicit-implicit-simulation-based optimization model yield good solutions to the sensor placement problem. A principal goal of the first part of the dissertation is to quantify the use of mobile and stationary sensors in incident detection.;The second part of the dissertation addresses a near future scenario, where hundreds of ITS sensors that have wireless networking capabilities are deployed in the road network system. To handle this large sensor network we propose the formation of clusters in the wireless sensor network. Cluster formation helps in network stabilization, avoids data duplication and avoids the formation of data bottlenecks. In the wireless sensor network literature, this problem is typically handled using rule based heuristics that select a clusterhead and form clusters. However, this approach, though practical, ignores the information regarding sensor movement. We use a mathematical programming based approach to clustering. The mathematical formulation developed for the clustering problem is NP-Hard. We propose heuristic solutions to find feasible initial solutions and improve the CPLEX performance using parameter tuning and warm starts. The mathematical formulation developed for the clustering problem is an approximate model. To understand the real-world behavior of the solution, we develop a simulation module that evaluates the sensor coverage possible through cluster formation.