01337241

Component

Predicted travel time information is important to successful implementation of many Intelligent Transportation Systems. An artificial neural network (ANN) approach to freeway travel time prediction is presented. However, one of the setbacks of ANN models is that many ANN topologies are developed following a “Black-box” approach, which is difficult to convey to many transportation professionals. Additionally, although input settings are one important factor that influences the ANN model performance, there are only a handful of studies focusing on the impacts of input information. In this study, we first provide some insights on how to logically use Predicted travel time information is important to successful implementation of many Intelligent Transportation Systems. The artificial neural network (ANN) is one advance approach to freeway travel time prediction. However, one of the setbacks of ANN models is that many ANN topologies are developed following a “Black-box” approach, which is difficult to convey the internal workings of these models to transportation practitioners. Additionally, although input settings are one important factor that influences the ANN model performance, there are only a handful of studies focusing on the impacts of input information. In this study, we first provide some insights on how to logically use knowledge about typical traffic processes to make the “White-box” oriented development of a neural network topology. We then employ a reliable ensemble technique to analyze the spatial and temporal effects of input variables on the ANN prediction performances for a study segment on US-290 in Houston. The results have shown that speed and occupancy data could be used by themselves or jointly to achieve satisfactory performance while traffic volume cannot; better performance can also be achieved by using inputs from upstream, current and downstream segments, and/or using inputs from current and one or two time steps in the past. At last, we utilize the understandings learned above to develop a new ANN topology, the so called time-delayed state-space neural network (TDSSNN). By comparing with other popular neural networks, the TDSSNN shows above-average prediction accuracy and consistency. But more importantly, the model illustrates the possibility of building a White-box ANN model.

01329018

11-2975

English

13p

2011

Transportation Research Board 90th Annual Meeting

DVD

Figures (4) ; References (34) ; Tables (4)

Advanced traveler information systems; Design; Network analysis (Planning); Neural networks; Traffic density; Traffic forecasting; Traffic models; Traffic speed; Traffic volume; Travel time

Houston (Texas)

Data and Information Technology; Highways; I71: Traffic Theory

Transportation Research Board Annual Meeting 2011 Paper #11-2975

TRIS, TRB

Feb 17 2011 6:21PM