PREDICTING INTERSECTION QUEUE WITH NEURAL-NETWORK MODELS

To capture the complex nature of intersection queue dynamics, this stu dy has explored the use of neural network models with data from extens ive simulation experiments. The proposed models, although lacking in m athematical elegance, are capable of providing the acceptable predicti on accuracy (more than 90%) at 3 time-steps ahead. As each time-step i s as short as 3 s, the resulting information on queue evolution is suf ficiently detailed for both responsive signal control and intersection operations. To accommodate the differences in available surveillance systems, this study has also investigated the most suitable neural net work structure for each proposed queue model with extensive explorator y analyses.