Arterial Traffic Activity Estimation
- Author(s): Yang, Qichi
- Advisor(s): Barth, Matthew
- et al.
With the advances in sensing technologies along with innovative modelling and estimation method, a variety of Intelligent Transportation Systems have been developed to mitigate traffic congestion and associated environment pollution problems. In recent decade, more advanced sensing technologies have been invented for dedicated measurements. In this dissertation, we studied traffic activity estimation algorithm with three types of advanced sensing systems: 1) sparse mobile sensors for arterial travel time estimation; 2) wireless magnetic sensors for arterial energy/emission estimation; and 3) 3-D lidar for lane-level vehicle trajectory estimation.
Due to the interruption by traffic signals, it becomes very challenging to estimate average travel time of traffic flows along the signalized arterial by using conventional inductive loop detectors (ILD). Position samples from fast-growing smart phone and commercial navigator technologies turn out to be another promising data source for this task. However, one of the major obstacles of using these technologies is the randomness of sampling location, which results in significant variation in measured distance between two consecutive samples, compared to the ILD technology. In Chapter 3, we developed a novel probabilistic travel time model to deal with this issue by decomposing the arterial travel time into two components: free-flow travel time and delayed time. Validated by field operational tests, the proposed model has exhibited a good fit on the travel time distribution under different congestion levels and has resulted in more reliable and robust vehicle's activity classification to differentiate stopped and free flow maneuvers by each individual vehicle. With the second benefit, we developed an arterial energy/emission estimation approach in Chapter 4, using wireless magnetic sensors which measure travel time directly for each re-identified vehicle. An approximated speed trajectory is then reconstructed for stopped vehicles and fed into a microscopic energy/emissions model to achieve more accurate energy/emissions estimation.
Lane-level second-by-second vehicle trajectory is another important data source, which is particularly useful for traffic simulators to calibrate the internal vehicle activity models. In Chapter 5, we present a novel mobile sensor platform consisting of centimetre-level accurate positioning system and 3-D lidar for surrounding vehicle trajectory collection. A robust detection/ tracking algorithm has been developed to extract a large number of trajectories from vehicles surrounding the sensor-equipped probe vehicle. Results from both freeway and arterial have shown great potentials of such innovative sensing systems in building high quality trajectory repository for future research.