A Hidden Markov Model for short term prediction of traffic conditions on freeways

Accurate short-term prediction of traffic conditions on freeways and major arterials has recently become increasingly important because of its vital role in the basic traffic management functions and trip decision making processes. Given the dynamic and stochastic nature of freeway traffic, this study proposes a stochastic approach, Hidden Markov Model (HMM), for short-term freeway traffic prediction during peak periods. The data used in the study was gathered from real-time traffic monitoring devices over six years on a 60.8-km (38-mile) corridor of Interstate-4 in Orlando, Florida. The HMM defines traffic states in a two-dimensional space using first-order statistics (Mean) and second-order statistics (Contrast) of speed observations. The dynamic changes of freeway traffic conditions are addressed with state transition probabilities. For a sequence of traffic speed observations, HMMs estimate the most likely sequence of traffic states. The model performance was evaluated using prediction errors, which are measured by the relative length of the distance between the predicted state and the observed state in the two-dimensional space. Reasonable prediction errors lower than or around 10% were obtained from HMMs. Also, the model performance was not remarkably affected by location, travel direction, and peak period time. The HMMs were compared to two naive predication methods. The results showed that HMMs perform better and are more robust than the naive methods. Therefore, the study concludes that the HMM approach was successful in modeling short-term traffic condition prediction during peak periods and in accounting for the inherent stochastic nature of traffic conditions. Published by Elsevier Ltd.