Pedestrian Navigation using Artificial Neural Networks and Classical Filtering Techniques

The objective of this thesis is to explore the improvements achieved through using classical filtering methods with Artificial Neural Network (ANN) for pedestrian navigation techniques. ANN have been improving dramatically in their ability to approximate various functions. These neural network solutions have been able to surpass many classical navigation techniques. However, research using ANN to solve problems appears to be solely focused on the ability of neural networks alone. The combination of ANN with classical filtering methods has the potential to bring beneficial aspects of both techniques to increase accuracy in many different applications. Pedestrian navigation is used as a medium to explore this process using a localization and a Pedestrian Dead Reckoning (PDR) approach. Pedestrian navigation is primarily dominated by Global Positioning System (GPS) based navigation methods, but urban and indoor environments pose difficulties for using GPS for navigation. A novel urban data set is created for testing various localization and PDR based pedestrian navigation solutions. Cell phone data is collected including images, accelerometer, gyroscope, and magnetometer data to train the ANN. The ANN methods are explored first trying to achieve a low root mean square error (RMSE) of the predicted and original trajectory. After analyzing the localization and PDR solutions they are combined into an extended Kalman Filter (EKF) to achieve a 20% reduction in the RMSE. This takes the best localization results of 35m combined with underperforming PDR solution with a 171m RMSE to create an EKF solution of 28m of a one hour test collect