Skip to Main Content
We develop an algorithm for predicting the arrival times of a transit vehicle at signalized intersections, with a focus on meeting the accuracy requirement associated with signal priority control applications. The algorithm uses both historical and real-time Global positioning system (GPS) vehicle location data. There are no data from other detectors, such as loops or cameras. The arrival time prediction is formulated as an optimal a posteriori parameter estimation problem, where the model is consisted of a historical model and an adaptive model that adaptively adjusts its filter gain based on real-time data. The estimates generated by these two models are fused in a weighted average derived from the solution of the parameter estimation problem. The prediction algorithm adaptively adjusts its weight distribution using error variances obtained from the two models. We include some simulations of field test results and their statistics to demonstrate the performance and convergence of the solution.