Predictive Traffic Operations and Control of Connected and Automated Vehicle Systems

Elfar, Amr

doi:https://doi.org/10.21985/n2-8hw8-h558

Work

Predictive Traffic Operations and Control of Connected and Automated Vehicle Systems

Public

Download PDF

Connected and automated vehicle (CAV) technology is a disruptive transportation development with potentially transformative impacts on society and the economy. CAV systems promise to significantly reduce human-caused road crashes, improve traffic flow performance, and lower pollutant emissions. However, realizing those benefits requires strategic planning for the deployment of CAV systems and developing advanced traffic control algorithms that utilize their new capabilities. To that end, the objective of this dissertation is to develop innovative traffic management strategies that utilize the big stream of data generated by CAV systems and the predictive capability of machine learning algorithms. The dissertation starts by introducing a methodological framework for developing predictive traffic management and control strategies utilizing CAV systems. It considers three main components: 1) traffic monitoring, 2) traffic state prediction, and) control strategy. Following this framework, the dissertation presents a novel method to identify shockwave formation and track its propagation based on the speed distribution of individual vehicles available through connected vehicles technology. The analysis shows a consistent pattern where shockwave formation, indicated by a speed drop propagating over space and time, is associated with a sharp increase in the value of speed standard deviation (SSD). Building on the aforementioned method, the dissertation also presents online and offline models for short-term traffic congestion prediction. Offline models are calibrated based on historical data and are updated (re-trained) whenever significant changes occur in the system, such as changes to the infrastructure. Online models are calibrated using historical data and updated regularly using real-time information on prevailing traffic conditions broadcasted by CAVs. Results show that the accuracy of the proposed models can reach 97%. Utilizing the early shockwave detection method and the congestion prediction models, the dissertation presents a predictive speed harmonization system with two CAV control strategies: centralized and decentralized. The centralized system relies on a traffic management center to collect data from CAVs within a road segment of interest, predict traffic congestion, and broadcast updated speed limits to CAVs in order to mitigate congestion. The decentralized system relies on individual CAVs to collect data through communicating with each other, predict traffic congestion using vehicle-specific models, and update their speed limits to mitigate congestion. Case studies of multiple operational scenarios show that both systems can reduce the severity and lengths of traffic shockwaves, improve the overall traffic stability, increase overall speed, and reduce travel time. The decentralized strategy can be more cost-efficient over the long run since it does not require any more resources beyond what CAVs are expected to have. For it be effective, however, cross-communication between different CAV fleets is required. Finally, the dissertation presents truck platooning as a special application of decentralized CAV traffic control strategies. Results show that forming platoons under an opportunistic strategy can be difficult due to the generally low number of trucks on highways.

Creator