Theses and Dissertations at Montana State University (MSU)
Permanent URI for this collectionhttps://scholarworks.montana.edu/handle/1/733
Browse
1 results
Search Results
Item Inference of passenger ridership, O-D flows, wait times, and travel times using Wi-Fi and GPS signals(Montana State University - Bozeman, College of Engineerng, 2019) Videa Martinez, Aldo Alejandro; Chairperson, Graduate Committee: Yiyi WangReal-time data collection of transportation parameters is a vital element of research and industrial applications. A real-time ridership data collection would facilitate the planning of trips and optimization of routes for transportation agencies. Riders also would have the ability to plan their trips more easily. With the advent of smartphone technologies, societies have obtained a new infrastructure that is based on wireless networks. This infrastructure can be used as a platform to obtain information. In our case, we are exploring the Wi-Fi networks that surround a space to obtain traffic data. Our research focuses on scanning IEEE 802.11 networks with a sniffing software and then we classified the different signals into passengers and not passengers. To do so, we used the various attributes obtained with our sniffing software on a Raspberry Pi computer and filtered the signals that would not belong to passengers. Afterward, we implemented machine learning algorithms on the pre-processed data to understand the intrinsic nature of the data and evaluate if there could be some traits that would be utilized to perform an unsupervised classification that corresponded to the passengers' smartphones. In the unsupervised learning algorithm, the parameters were reduced into two with arithmetic operations and principal component decomposition. The results with the rule-based methodology are more accurate than the unsupervised methodology. We believe this is due to many router signals that are similar to the passengers' smartphones. Our proposed methodology has some limitations like some riders not carrying smartphones, and overestimation resulted from the noise of other signals. However, with the devices that were detected, we have demonstrated that the time of detection is accurate which helped us infer the origin-destination flows from a portion of our subjects. Additionally, we used the GPS traces to estimate travel time between buses and wait time of passengers at a bus stop.