LORITA ANGELINE et al: TRAFFIC JAM CLUSTERING ANALYSIS WITH COUNTERMEASURE STRATEGIES . . DOI 10.5013/IJSSST.a.21.02.04 4.1 ISSN: 1473-804x online, 1473-8031 print Traffic Jam Clustering Analysis with Countermeasure Strategies During Traffic Congestion Lorita Angeline, Renee Ka Yin Chin, Chung Fan Liau, Ismail Saad, Kenneth Tze Kin Teo* Modelling, Simulation & Computing Laboratory, Artificial Intelligence Research Unit Faculty of Engineering, Universiti Malaysia Sabah, Kota Kinabalu, Malaysia. * Corresponding author: ktkteo@ieee.org e-mail: angeline.lorita@gmail.com, reneekychin@ums.edu.my, cfliau@ums.edu.my, ismail_s@ums.edu.my Abstract- Countless hours are lost in traffic jam every year. In the efforts to save time, drivers tend to speed up in traffic jam. However there is a traffic paradox suggests that speeding up in traffic jam may not necessarily time saving. Traffic flow is fundamentally dynamic in nature, where the flow formed is greatly subjected to the interaction amongst the drivers. As such, this paper aims to investigate how different speed weightage between the drivers instigating the jam clusters and to assess possible corrective action to reverse jam clusters formation. Based on the identified jam clusters, several properties such as cluster lengths and average speed within the clusters are analysed for corrective action. Traffic simulation with 300 samples on a 5 km length of road shows the proposed algorithm does improve travel time with improvement range from 1.96 % to 7.96 %. Keywords - traffic modelling; jam clustering; traffic cellular automata; stop-and-go traffic waves; TCA I. INTRODUCTION Does speeding get the drivers to their destination faster? The traffic paradox ‘faster is slower’ may actually valid at certain circumstances [1]. This paper is inspired by [2, 3], which are the first field experiments reporting that traffic waves can materialize as a result of human driving behaviour alone. These experiments were an eye opening to many researchers, but they do not offer a justification in reversing the traffic waves. To address this gap, this paper aims to investigate how different speed values cause the stop-and-go traffic waves and to assess possible corrective action to reverse the jam clusters formation. Thus travel times are selected as objective function. This paper is organized as follows: Section II presents the cellular automata (CA) and the settings for the traffic platform. Section III discusses the jam clusters formation and their properties. Section IV presents a possible corrective action algorithm, simulation results, and analysis. The overall project is concluded in Section V. II. TRAFFIC CELLULAR AUTOMATA MODELLING The implementation of CA can be traced to the early 1950’s, with the most renowned NaSch traffic cellular automata (TCA) model introduced in 1992 [4, 5]. NaSch model is capable to imitate traffic nature and reproduced the emergence of traffic jam. The difference among TCA models primarily in the update rule. Generally there are four rules in NaSch model: acceleration, deceleration, randomisation and advancing [6]. Traffic researchers acknowledge NaSch model as a minimal model, in the context that all these rules are the essential for mimicking the fundamental components in traffic flow. According to [5], the randomisation rule emulates natural speed fluctuation due to driver behaviour. Although the justification is widely agreed upon, much disapproval was however expressed due to the randomisation rule [7, 8]. Nevertheless, NaSch model has served as traffic platform for decades, credit to its versatile algorithm that made tweak and adjustment on the model rather swiftly. Therefore, there are other adaptations of NaSch model, each with various levels of aim and objective function such as the study of fuel and energy dissipation in traffic stream [9, 10], modelling driver behaviour [11 – 13] and effect of acceleration and deceleration in the traffic stream [14, 15]. Thus, the proposed traffic model in this paper is developed based on NaSch traffic model, with adaptations on the acceleration and deceleration rules. Table I shows the settings for the traffic CA model [6]. The next section analyses the emergence of jam clusters simulated from the proposed traffic model. TABLE I. TRAFFIC CA MODEL PRELIMINARY SETTINGS Variables Settings Type of CA Stochastic Boundary condition Open Average vehicle length 5 m Occupied space (cell’s width) 7.5 m Driver’s reaction time 0.75 sec Speed limit 110 km/h III. JAM CLUSTERS ANALYSIS The commonly used parameters such as traffic volume, density and speed may not enough to determine the state of traffic congestion. Speed of zero can be caused by severe traffic jam (gridlock) or it could be no vehicles on road.