In recent years we have seen a significant increase in research on autonomous vehicles, which is attracting considerable attention because of its potential to revolutionize transportation. In this paper, a method is presented for classifying static and dynamic tracks in autonomous vehicles using odometry and LiDAR data. Using the multi-target tracking algorithm, we are able to detect and track objects with high precision. We used an unsupervised clustering algorithm to create innovative interaction models, which will be employed for vehicle localization in the future. Because clustering is based upon time and spatial proximity, we are able to create interaction models using dynamic bayesian networks (DBNs) that allow us to classify and differentiate static from dynamic tracks. Based on our method, we achieve an accuracy of 87% in distinguishing between static and dynamic obstacles, thereby enhancing navigation by identifying and differentiating different types of obstacles efficiently.