Bibliographic citations
Valdivia, H., (2020). Estudio de predicción de flujo de tráfico a corto plazo basado en el filtro de Kalman y la transformación Wavelet [Tesis, Universidad Nacional de Ingeniería]. http://hdl.handle.net/20.500.14076/21724
Valdivia, H., Estudio de predicción de flujo de tráfico a corto plazo basado en el filtro de Kalman y la transformación Wavelet [Tesis]. PE: Universidad Nacional de Ingeniería; 2020. http://hdl.handle.net/20.500.14076/21724
@misc{renati/711368,
title = "Estudio de predicción de flujo de tráfico a corto plazo basado en el filtro de Kalman y la transformación Wavelet",
author = "Valdivia Huerta, Henry Christhian",
publisher = "Universidad Nacional de Ingeniería",
year = "2020"
}
Short-term traffic flow prediction is an important component for proactive application development in Intelligent Transportation Systems (ITS). It is especially important for traffic management and control systems since these improve the efficiency and safety of traffic operations, but due to the stochastic nature of the process that governs traffic flow, it is a challenging task to find prediction algorithms that are accurate and robust. The task is further complicated by the lack of data availability in much of the transport systems of developing countries. This thesis proposes and studies a short-term traffic flow prediction method based on the Kalman filter and the discrete wavelet transform with a limited database of two days of observation. Discrete wavelet analysis using the coefficient thresholding technique helps filter or reduce noise contained in the data and the Kalman filter projects traffic flow onestep ahead based on the filtered data. Both filtering and projection processes are integrated in a single recursive real time process. The present thesis investigates the influence on the prediction accuracy of 1) different types of coefficients that can be integrated by the system state vector, 2) the type of Kalman filter used (conventional or adaptive), and 3) activation (true or false) of the noise filtering within the prediction process. In total, there are three parameters whose variation together generates twelve prediction models. Prediction models evaluation were performed under normal and disruptive traffic conditions. It is concluded from the research that the noise filtering process significantly improves the prediction performance under normal traffic conditions. However, this worsens under disruptive conditions. Prediction models based exclusively on the Adaptive Kalman filter are the most robust of the set of models proposed in face of any kind of traffic condition and their prediction accuracy varies according to the type of state vector used.
This item is licensed under a Creative Commons License