DISPATCHING AND RELOCATION OF EMERGENCY VEHICLES ON FREEWAYS: THEORIES AND APPLICATIONS

Resource allocation decisions are made to serve the current emergency without knowing which future emergency will be occurring. Different ordered combinations of emergencies result in different performance outcomes. Even though future decisions can be anticipated with scenarios, previous models follow an assumption that events over a time interval are independent. This dissertation follows an assumption that events are interdependent, because speed reduction and rubbernecking due to an initial incident provoke secondary incidents. The misconception that secondary incidents are not common has resulted in overlooking a look-ahead concept. This dissertation is a pioneer in relaxing the structural assumptions of independency during the assignment of emergency vehicles. When an emergency is detected and a request arrives, an appropriate emergency vehicle is immediately dispatched. We provide tools for quantifying impacts based on fundamentals of incident occurrences through identification, prediction, and interpretation of secondary incidents. A proposed online dispatching model minimizes the cost of moving the next emergency unit, while making the response as close to optimal as possible. Using the look-ahead concept, the online model flexibly re-computes the solution, basing future decisions on present requests. We introduce various online dispatching strategies with visualization of the algorithms, and provide insights on their differences in behavior and solution quality. The experimental evidence indicates that the algorithm works well in practice. After having served a designated request, the available and/or remaining vehicles are relocated to a new base for the next emergency. System costs will be excessive if delay regarding dispatching decisions is ignored when relocating response units. This dissertation presents an integrated method with a principle of beginning with a location phase to manage initial incidents and progressing through a dispatching phase to manage the stochastic occurrence of next incidents. Previous studies used the frequency of independent incidents and ignored scenarios in which two incidents occurred within proximal regions and intervals. The proposed analytical model relaxes the structural assumptions of Poisson process (independent increments) and incorporates evolution of primary and secondary incident probabilities over time. The mathematical model overcomes several limiting assumptions of the previous models, such as no waiting-time, returning rule to original depot, and fixed depot. The temporal locations flexible with look-ahead are compared with current practice that locates units in depots based on Poisson theory. A linearization of the formulation is presented and an efficient heuristic algorithm is implemented to deal with a large-scale problem in real-time.

SEQUENTIAL DECISIONS AND PREDICTIONS IN NATURAL LANGUAGE PROCESSING

Natural language processing has achieved great success in a wide range of ap- plications, producing both commercial language services and open-source language tools. However, most methods take a static or batch approach, assuming that the model has all information it needs and makes a one-time prediction. In this disser- tation, we study dynamic problems where the input comes in a sequence instead of all at once, and the output must be produced while the input is arriving. In these problems, predictions are often made based only on partial information. We see this dynamic setting in many real-time, interactive applications. These problems usually involve a trade-off between the amount of input received (cost) and the quality of the output prediction (accuracy). Therefore, the evaluation considers both objectives (e.g., plotting a Pareto curve). Our goal is to develop a formal understanding of sequential prediction and decision-making problems in natural language processing and to propose efficient solutions. Toward this end, we present meta-algorithms that take an existent batch model and produce a dynamic model to handle sequential inputs and outputs. Webuild our framework upon theories of Markov Decision Process (MDP), which allows learning to trade off competing objectives in a principled way. The main machine learning techniques we use are from imitation learning and reinforcement learning, and we advance current techniques to tackle problems arising in our settings. We evaluate our algorithm on a variety of applications, including dependency parsing, machine translation, and question answering. We show that our approach achieves a better cost-accuracy trade-off than the batch approach and heuristic-based decision- making approaches. We first propose a general framework for cost-sensitive prediction, where dif- ferent parts of the input come at different costs. We formulate a decision-making process that selects pieces of the input sequentially, and the selection is adaptive to each instance. Our approach is evaluated on both standard classification tasks and a structured prediction task (dependency parsing). We show that it achieves similar prediction quality to methods that use all input, while inducing a much smaller cost. Next, we extend the framework to problems where the input is revealed incremen- tally in a fixed order. We study two applications: simultaneous machine translation and quiz bowl (incremental text classification). We discuss challenges in this set- ting and show that adding domain knowledge eases the decision-making problem. A central theme throughout the chapters is an MDP formulation of a challenging problem with sequential input/output and trade-off decisions, accompanied by a learning algorithm that solves the MDP.