Prediction on Travel-Time Distribution for Freeways Using Online Expectation Maximization Algorithm

This paper presents a stochastic model-based approach to freeway travel-time prediction. The approach uses the Link-Node Cell Transmission Model (LN-CTM) to model traffic and provides a probability distribution for travel time. On-ramp and mainline flow profiles are collected from loop detectors, along with their uncertainties. The probability distribution is generated using Monte 5 Carlo simulation and the Online Expectation Maximization clustering algorithm. The simulation is implemented with a reasonable stopping criterion in order to reduce sample size requirement. Results show that the approach is able to generate an accurate multimodal distribution for traveltime. Future improvements are also discussed.