Ghassan Abu-Lebdeh - Department of Civil Engineering, American University of Sharjah
Timothy J. Likens - Well + Associates, Novi, Michigan, USA

Short term travel time prediction on urban arterials is an important component of Advanced Traveler Information Systems (ATIS) and Advanced Traffic Management Systems (ATMS). It can also be a key input to evacuation and emergency responses. This study presents robust travel time prediction models that work efficiently for time-variant turn movement shares in both congested and non-congested conditions that urban arterials typically experience throughout the day and/or during special events. The state-space notion of traffic processes was found useful and State-Space Neural Network models are proposed. Models were developed for travel time prediction on links, arterials, and routes (multiple links on different arterials). Mean absolute percentage errors of modeled travel times on arterilas for through, left, and right movements ranged between 12.3% and 34.6% for testing data sets. For routes, the MAPE ranged between 8.5 and 10%.

State-space Neural Networks, Travel Time, Urban Arterials