STOCHASTIC TRAFFIC DEMAND PROFILE

Abstract

Traffic surveys routinely estimate the profile of traffic demand in a certain road section, showing the expected evolution of the demand over a workday or weekend. However, the actual demand fluctuates around this value. That can lead to brief excess of the capacity at the moment of high demand and consequent congestion due to the capacity drop. This type of traffic demand variability has not yet been properly studied despite the fact it can play significant role in traffic modelling and engineering applications. This paper presents results of analysis of demand variability in five-minute aggregation intervals. The results do not clearly show a single random distribution that would accurately model the demand variability. Normal, lognormal and gamma distributions all show reasonably well fit to the data for individual intervals. Based on count of best fits, the lognormal distribution seems best, but in most cases, the difference between the distributions is not statistically significant. There appears to be a pattern where certain distributions have better fit in different times of day and week. The regularity and magnitude of demand (e.g. morning peak hour) probably play a role in this, as well as the aggregation interval.