Real-time traffic state estimation plays an imperative role in traffic management. For homogenous traffic, where all vehicles are the same type, traffic state estimation utilising parameters like speed, flow, density, and queue length is appropriate. Due to the disparities in static and mechanical or kinematic characteristics among vehicles and the asymmetric interactions that ensue, these are inadequate for mixed traffic conditions. Hence, the study's first objective proposes a speed-based characterization methodology for mixed traffic. This objective develops class-specific speed-based characterization through novel traffic state definition and ordering of the traffic states from free-flow to congested conditions are based on area occupancy. Parametric and non-parametric prediction models are also constructed for state and class-wise speed prediction. The results demonstrate that joint models used in state prediction are superior and computationally efficient than marginal models used in class-wise speed prediction.
Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication systems are an integral part of the Intelligent Transportation Systems. In this system, the probe vehicle continuously collects traffic information about the surrounding vehicles using onboard sensors, such as radar, LiDAR, cameras, GPS and WiFi sensors. This data includes information about the vehicle's speed, position, and acceleration, These data are transmitted to the roadside infrastructure using dedicated short-range communication (DSRC) technology and then sent to the remote server using General Packet Radio Service (GPRS) communication. Thus, in the second objective, through quasi-connected vehicle systems (V2V, V2I), higher quality and quantity of data from surrounding vehicles are used for local traffic state estimation along the path of the mobile sensor. In this objective, two possible scenarios are considered for state estimation: i) Estimation of traffic state with limited surrounding awareness. and ii) Estimation of traffic state with complete surrounding awareness. Limited surrounding awareness refers to the availability of approximate position data of sample surrounding vehicles with vehicle type information remaining unknown. On the other hand, complete surrounding awareness refers to the availability of all the surrounding vehicles' position and vehicle type information. Two different traffic state estimation methods are proposed for scenario 1 (limited surrounding awareness). Method 1 is based on the three-dimensional extension of Edie's definitions, and Method 2 is based on the cumulative number of vehicles (N). Similar to scenario 1, two different methods are proposed for scenario 2 (complete surrounding awareness). Method 1 utilises the speed-based characterization of objective 1 using the vehicle type information. The second method of scenario 2 is based on vehicle-area-based state variables defined for mixed traffic conditions.
To extend this further to the continuous spatio-temporal traffic state estimation along corridors, the third objective utilizes both stationary and mobile sensors. Extending the `Three-detector problem' framework for cumulative area curves facilitates the spatio-temporal state estimation of areal flow, areal density and speed feasible. Thus, this study develops novel contributions for characterization and state estimation for contemporary and connected vehicles in mixed traffic leveraging stationary and mobile sensor data.