Optimization of traffic simulation using GPS navigation records
MetadataShow full item record
A traffic simulation is a tool that permits constructing a virtual environment based on a real one, with the objective to perform analysis about the actual conditions and more important, apply changes to the virtual scene or to the driving rules to generate new scenarios and test solutions. However, the problem we found with simulations it that with incorrect parameters may not represent the traffic conditions we are looking for. In this work, we propose a method to calibrate the traffic simulations using data available for transportation in Costa Rica. This data comes from Global Position System (GPS) navigation records. The calibration algorithm search to represent those actual traffic conditions in a virtual environment, and after that, propose and design solutions to ease the complicated traffic situations. This thesis reflects the work of months to design and implemented an algorithm to calibrate simulations of five sectors of the country where we found difficult traffic conditions. The algorithm calculates a Measure of Performance to compare data from the simulation and the GPS records, and it searches iteratively for the best parameters. In the end, it validates the best solution found with a statistical test. As results, we achieved to calibrate the simulations for the five studied sectors, reaching a configuration of input parameters that reflects the traffic conditions extracted from the GPS records, as a portrait of the real-life conditions of the locations. The impact and applications of this work are plenty. For the computing part, we can dig more profound in using more techniques of calibration, and also exploit the data available for more general works. Moreover, it can become in a significant resource for analysis and decision making in urban mobility studies.
Proyecto de Graduación (Maestría en Computación con énfasis en Ciencias de la Computación) Instituto Tecnológico de Costa Rica, Escuela de Ingeniería en Computación, 2021