A Reliable Travel Time Prediction System With Sparsely Distributed Detectors

摘要: Title of Document: A RELIABLE TRAVEL TIME PREDICTION SYSTEM WITH SPARSELY DISTRIBUTED DETECTORS Nan Zou Directed By: Dr. Gang-Len Chang, Professor Department Civil and Environmental Engineering Due to the increasing congestion in most urban networks, providing reliable trip times commuters has emerged as one critical challenges for all existing Advanced Traffic Information Systems (ATIS). However, predicting travel time is a very complex difficult task, resulting accuracy varies with many variables time-varying nature, including day-to-day traffic demands, responses individual drivers daily commuting congestion, conditions road facility, weather, incidents, reliability available detectors. This study aims develop prediction system that needs only small number detectors perform accurate real-time predictions under recurrent conditions. To ensure its effectiveness, proposed consists three principle modules: estimation module, missing data module. The module specially designed hybrid structure responsible estimating scenarios or without sufficient field observations, supplying estimated results support developed take full advantage various information, historical times, geometric features, daily/weekly patterns. It can effectively deal patterns multiple embedded models, primary multi-topology Neural Network model rule-based clustering function supplemental an enhanced k-Nearest Neighbor model. contend issue, which occurs frequently any realworld system, this incorporates based on imputation technique estimate both shortand long-term so avoid interrupting operations. been implemented from 10 roadside 25-mile stretch I-70 eastbound, performance tested against actual collected by independent evaluation team. Results extensive have indicated capable generating types outperforms state-of-the-practice state-of-the-art models literature. Its also top methods are able maintain state when at key location experience rate 20% 100% during uncongested, congested transition periods.