Route Planning with Statistical Models

Abstract

One difficulty to find the fastest route in route planning is how to determine the precise travel time on each road. In the real world, the travel time of each road varies with time, weather condition and many other factors. The thesis aims at studying route planning algorithms that use statistical models to predict the changes of travel time for each road and calculate the fastest route. Using the historical data of main roads in Washington D.C. area, the thesis studied major factors that would affect the travel time. Different statistical models are presented and compared to fit the travel time of each road. Then the LASSO regression model is chosen, and different predictive route planning algorithms are introduced to fulfill our goal. Finally, deterministic approximate dynamic programming is recommended to solve our problem.