Travel time reliability on a highway network: estimations using floating car data

Abstract

With the substantial increase in traffic in many urban areas, travel time reliability is becoming a more critical and more relevant factor than travel time. Currently, new indicators involving travel times and their variability are being used to better assess the efficiency of road networks. Our research here is an attempt to assess the reliability of travel times on the main highway corridors of the Montreal Area (Canada), using floating car data gathered from 1998 to 2004 by MTQ (Quebec's Ministry of Transport). This paper presents the outputs of the data analysis and modeling process that was developed to estimate travel times using such data, as well as to assess the level of variability of those times. Almost 30,000 travel time observations were gathered on fifty different routes over a 6-year period. These routes were divided into 1-kilometer road segments, which were analyzed and modeled using various techniques. The process involves finding the best statistical model to describe travel time distribution, while controlling for a number of factors (period, month, year, or weather) and identifying segments presenting high variability. Secondly, the mean time and time variability are simulated all along the routes and areas that suffer recurrent congestion. Finally, the analysis introduces two new indicators: the probability of non recurrent incidents, and an index summarizing both the mean travel time and the variability of travel times. In the future, we expect to be able to simulate the expected travel time per portion of a route, its reliability, and the probability of encountering incidents of any kind.

Keywords:

• Travel time
• Freeways
• Traffic congestion
• Floating cars