Evaluation of outlier filtering algorithms for accurate measurement of travel time reliability incorporating lane-splitting situations / Obada M. A. Asqool

Obada , M. A. Asqool (2022) Evaluation of outlier filtering algorithms for accurate measurement of travel time reliability incorporating lane-splitting situations / Obada M. A. Asqool. Masters thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (206Kb)
	PDF (Thesis M.A.) Download (1635Kb)

Abstract

Most developed and developing countries, including Malaysia, experience high traffic congestion, especially in urban areas. When congestion occurs, traffic moves at a lower speed which increases travel time and in consequence, people spend considerable time fulfilling their daily journeys. Measuring Travel Time Reliability (TTR) helps traffic professionals to quantify congestion based on travel time parameter, thus, adopting suitable strategies to mitigate traffic congestion. On the other hand, motorcycles are significantly high in Malaysia and other Association of Southeast Asian Nations (ASEAN) countries. As such, detection of travel time based on the media access control (MAC) address is not straightforward. Raw travel times data does not represent the actual traffic condition of passenger cars since motorcycles travel faster than cars during congestion using the gap between two parallel rows of cars. This situation is called lane-splitting. In the past, many travel time filtration algorithms were established. However, there is a real need to determine which algorithms can produce the most accurate results when considering actual and large datasets from lane-splitting situations. Therefore, this study aims to investigate the best algorithm for data filtration and how to use it to obtain accurate data for measuring TTR. In order to find the best filtration algorithm, two stages were adopted in this study. The first stage was the validation of the performance of the previous algorithms. The assessment was conducted by observing the performance of each algorithm and comparing its performance with other algorithms when lane-splitting data were applied. The second stage was to investigate the sensitivity of the algorithm parameters for different days. To analyse TTR, Travel Time Index (TTI), Planning Time Index (PTI), and Buffer Time Index (BTI) were calculated with respect to the time of day (TOD), day of week (DOW), holidays, and election days. This study used travel time datasets collected from three routes in Kuala Lumpur via Wi-Fi detectors during May 2018. The results showed that the Jang algorithm was found to have the best performance for two of the three routes, whereas the TransGuide algorithm was the best algorithm for one route. The parameters of the Jang algorithm and TransGuide algorithm were sensitive for different days. Thus, the Jang algorithm and TransGuide algorithm could be used after calibrating their parameters for each day. After the filtration, TTR measures were calculated. The analysis of TTR measures showed that on weekdays and weekends, the three routes suffered from high variability in travel time. On election days and holidays, the road network operates near to free-flow condition for most of the time with low variability in travel time. The findings and contribution of this research are beneficial for transportation companies and authorities that depend on MAC addresses to collect travel time data in Malaysia and ASEAN countries. Furthermore, this research introduced the concept of TTR, demonstrating its importance to Malaysian traffic researchers. Accordingly, Malaysian transportation authorities need to adopt TTR measures in their studies, reports, and guidelines to maintain reliable travel time.

Actions (For repository staff only : Login required)