Regular Vehicle Spatial Distribution Estimation Based on Machine Learning

For the mixed traffic flow, obtaining distribution of connected vehicles (CVs) and regular (RVs) is great significance for road network analysis cooperative control in intelligent transportation systems (ITSs). However, whether it based on fixed sensors or CVs mechanism to estimate spatial RVs, implementation complexity low estimation accuracy are points that need be improved. This paper proposes a vehicle method using adjacent as mobile sensors. First, investigate hidden relationship between interaction information RVs among CVs, Gaussian mixture model-hidden Markov model (GMM-HMM) selected identification method. Then, three sets experiments were designed study influence observed features capability model, generalization validation, comparison with other methods, respectively. Finally, proposed verified by dataset generated car-following model. The experimental results show selecting relative position time headway can effectively reflect CVs. average identify over 93.7%, which provide valuable suggestions Internet Vehicles application.