As pavement condition grows to be one of the crucial problems facing our national highway system, a new challenge emerges in developing pavement deterioration prediction models that are reliable yet easily applicable by Highway Pavement Management System (HPMS) in State DOTS and other agencies. This reports presents the research done in this area in Rutgers University. The significant contribution of this research lies in the fact that it utilizes the most comprehensive database of pavement conditions (LlTP) that is readily available and promises to provide the sought data in future years. The Long Term Pavement Project (LTTP) Database developed by the Federal Highway Administration was chosen to provide the required data of related parameters for the model development. The first part of this report reviews the existing literature covering related topics including pavement roughness, the Long Term Pavement Project LTPP background, artificial neural networks, regression analysis and the existing pavement deterioration models developed by Federal Highway Agency or reported by Transportation Research Record as well as the default model that is utilized by the Pavement Management System. The second part discusses the work done in data analysis and data manipulation in addition to the development of the training of the neural network model. The third part deals with various aspects of the model development using neural networks and regression analysis. The next part concludes the research with summarizing the results of model development and then by presenting a comparison between the models developed in this research and some existing models by applying these models to similar data sets and performing statistical analysis of the results. Lastly, the report presents some recommendations for future research in this area.