This study aims to develop an algorithm to estimate roadway congestion by deploying sophisticated artificial intelligence techniques on information collected from radar sensor vehicles. In our previous research, we designed an algorithm to estimate road density based on deep learning methods for a single continuous link, assuming the availability of data collected from sensor vehicles. The simulation results show that deep learning methods are better than the analytical density estimation methods developed through mathematical derivation, in an earlier phase of the project. In particular, due to the nature of the information collected through the sensor vehicles, the deep learning approach has a higher density estimation power than existing algorithms which have high estimation errors during the congestion onset and clearing periods. Previous studies, however, have not considered the traffic condition correlation between successive road segments, which is an important feature of traffic flow. In this study, we will develop an algorithm that accounts for the correlation between adjacent links, and design a deep-run model that can improve the traffic density estimation ability by incorporating the characteristics of links and traffic flows.  The primary indicators of traffic flow forecasting are not only traffic density but also traffic speed and traffic volume, which are closely related to each other.   In this study, we will develop a model that can predict these three indicators at the same time by using information obtained from sensor vehicles and deep learning analytics.