Short-term Travel Time Prediction with Neural Network Models Wonjae Jang Civil & Environmental Engineering ECE 539 Term Project
Project Description Application of ANN to predict travel time 1-5 time periods ahead for one roadway segment on Beltline highway (3-15 minutes prediction) Simulation of traffic condition using CORSIM Comparing results from several models run
Background Traffic Control Induce Traffic divergence Ramp metering Advance Traveler Information System (Route Guidance System) Need to predict travel time several time steps ahead Data Availability Surveillance systems (CCTV, detector) Other technologies
Existing Approaches Use Historical Data Use Real Time Data Historical + Real Time (aH(h,t)+(1-a)Rt(t-1)) ARIMA Kalman Filtering ANN
ANN Applied MLP – Back Propagation algorithm was used ANN I Use L u (t-1), L t (t-1), L d (t-1) as input ANN II Use L u (t-5),…, L u (t-1), L t (t- 5),…, L t (t-1), L d (t-5),…, L d (t-1) as input ANN III Use L 1 (t-1), L 2 (t-1),…, L n (t- 1), L u (t-1), L t (t-1), L d (t-1) w/wo R 1 (t-1), R 2 (t-1),…, R n (t-1) utd utd1…n 1 n
Results