Institute of Visual Informatics

Institute of Visual Informatics (IVI)

Leading Digital Technology Across Industrial Revolution

Predicting Traffic Flow Propagation Based On Neighbouring Roads Using Hidden Markov Model(HMM)

Headed by Assoc. Prof. Dr. Azlina Binti Ahmad.

Skim Geran Penyelidikan Fundamental (FRGS)



Traffic congestion in urban areas has become worse due to increase in population and number of private cars. Traffic congestion does not only lead to economic losses, but can also cause human stress, casualties due to accidents and environmental pollution. Thus, it is of high importance to study traffic congestion and its mitigation. Various studies in traffic flow have been conducted, but study on predicting traffic flow propagation is still limited. Monitoring traffic congestion is complex because it is dynamic. Traffic congestion can propagate from one road to neighbouring roads. It is therefore the main aim of the study is to propose and validate a model to predict traffic flow propagation based on neighbouring roads. Since traffic flow is a temporal process, we propose to model the propagation using Hidden Markov Model (HMM). HMM consists of a finite state set of states, each with a probabilistic function representing a hidden Markov chain. Previous studies focused on prediction of traffic condition on a road segment without taking into consideration its impact on neighbouring roads. For our study, we need to determine the road which will be used as a predicting factor for traffic flow propagation on neighbouring roads. Parameters that we will consider include distance between road segments, traffic direction, time and day which can be represented by the sequence Xr= {x(S,t)│t=1,2,….,M} where S represents the road segment. We will adopt a clustering algorithm to cluster time and day to obtain similarity pattern between the road segments. The methodology comprises of investigation on impact of traffic flow on neighbouring roads, development and evaluation of model. The expected output of this study is a HMM based model to predict traffic flow propagation. This model can be implemented in any application for predicting traffic flow propagation for road users.