Highway Lane-Changing Prediction Using a Hierarchical Software Architecture based on Support Vector Machine and Continuous Hidden Markov Model

Lane changing behavior is one of the most essential and complex driving attributes. The lack of proper lane changing behavior can lead to collisions and traffic congestion. In this work, a novel hierarchical software architecture for the prediction of lane changing behavior on highways has been developed and evaluated. The two-layer hierarchical structure of the proposed model is based on a support vector machine (SVM) in the first layer followed by another model based on continuous Hidden Markov Model (HMM) incorporated with a Gaussian Mixture Model (GMM) in the second layer. The trajectory classification predicted in the first layer by the SVM is binary, i.e., Lane Change (LC) and Lane Keep (LK) behaviors. The second layer of the software architecture further classifies the LC behavior output of the first layer to left-lane change (LLC) and right-lane change (RLC) behaviors using the model of continuous HMM (CHMM) incorporated with GMM. The developed model has been evaluated using the real-world dataset of U.S. Highway 101 and Interstate 80 from Federal Highway Administration’s Next Generation Simulation (NGSIM). The first layer prediction is performed within an approximately 10 seconds time window. The positions, velocity and Time to Collision (TTC) of the target and surrounding vehicles are taken as input parameters in the model execution of the second layer. The test results show that the proposed hierarchical model exhibits 91% accuracy for LLC, 87% accuracy for RLC and 99% accuracy for LK behaviors. This model can be effectively used as a lane changing suggestion system in the advanced driver assistance systems (ADAS).