An optimization framework for hazardous materials route selection problem

Hazardous Materials (hazmats) play a, crucial role in important areas of many industrial sectors such as heavy metal industries, agriculture, major pharmaceutical industries, oil and fuel industries, and many others. Hazardous Materials Incident Reporting System reported approximately 15,000 hazmat incidents every year in the United States. The occurrences of hazmat incidents although very low; pose severe dangers to the environment and community. We first develop a series of statistical models that help predict the expected cost of a hazmat incident. Further, we show that the transformation of a linear regression model has the ability to capture a wide range of incident costs. Using these data, we develop a Markov Decision Process model to plan routes for hazmats with a single objective and then propose a generalized model with multiple L(L epsilon Z+), objectives. Our research work further incorporates the variability of factors affecting the route selection by time of the day. This problem was solved using dynamic programming but is computationally expensive. We proposed three solution techniques for this problem. We show that two of such techniques compute optimal routing policies and expedite the solution process by at least 50-60%. Lastly, we address a new problem related to this routing known as the information based routing problem. We propose a web-enabled framework for generating routes in real time for trucks by taking into account route conditions.