A Study on Self Similar Vehicle Arrival Pattern at Isolated Signalized Intersection

Chew, Carl Jun (2018) A Study on Self Similar Vehicle Arrival Pattern at Isolated Signalized Intersection. Master dissertation/thesis, UTAR.

Preview

PDF Download (1987Kb) | Preview

Abstract

In data network, network arrivals are modeled based on Poisson assumptions but later it was discovered that Local Area Network (LAN) traffic possesses self-similar characteristics (Leland et al., 1994). Paxson and Floyd (1995) discovered that Wide Area Network (WAN) traffic is well modeled by using self-similar process rather than Poisson process. Since then, self-similar processes have been studied and implied in data network. Conventionally, the vehicle arrival pattern is also modeled based on Poisson assumptions in traffic studies. In traffic signal design, vehicle arrival pattern is important in calculating the queue length and cycle length. However, recent researches have shown that Poisson assumptions no longer hold under moderate to heavy traffic conditions. Nagatani (2005) discovered that individual vehicle that passes a series of traffic lights possessed self-similar characteristics. Meng and Khoo (2009) concluded that the vehicle arrival pattern on highway possessed self-similar characteristics. These recent studies have shown that vehicle arrival patterns possess self-similar characteristics. Therefore Poisson assumptions made on vehicle arrival pattern should be reconsidered and the existence of self-similar characteristics in vehicle arrival pattern should be examined. This research aims at investigating the existence of self-similarity characteristics for vehicle arrival pattern and its impact on traffic signal design. This research emphasizes on the vehicle arrival pattern of isolated signalized intersections within the Kuala Lumpur City Centre. By recording the movement of incoming vehicles, the vehicle arrival patterns and its corresponding time headway is tabulated and analyzed. Then, statistical analysis such as hypothesis tests was executed to examine the goodness-of-fit between Poisson process and self-similar process. It is discovered that the vehicle arrival pattern in these intersections exhibit self-similarity characteristics and its corresponding time headway distribution is heavy-tailed. By affirming the self-similarity characteristics in vehicle arrival pattern, this thesis developed a new approach in calculating the average delay of the vehicles under different values of Hurst parameter, which is an indicator for self-similarity. Then, the formulation of queue length and cycle length are derived based on self-similar characteristics. It is aimed that this new approach can provide a better and more accurate alternative in the construction of traffic signal design to optimize the average delay of the vehicles. This is crucial in providing a more precise and accurate queue length and cycle length for traffic signal design.

Actions (login required)