Connected and Autonomous Vehicles Applications Development and Evaluation for Transportation Cyber-Physical Systems

Cyber-Physical Systems (CPS) seamlessly integrate computation, networking and physical devices. A Connected and Autonomous Vehicle (CAV) system in which each vehicle can wirelessly communicate and share data with other vehicles or infrastructures (e.g., traffic signal, roadside unit), requires a Transportation Cyber-Physical System (TCPS) for improving safety and mobility, and reducing greenhouse gas emissions. Unfortunately, a typical TCPS with a centralized computing service cannot support real-time CAV applications due to the often unpredictable network latency, high data loss rate and expensive communication bandwidth, especially in a mobile network, such as a CAV environment. Edge computing, a new concept for the CPS, distributes the resources for communication, computation, control, and storage at different edges of the systems. TCPS with edge computing strategy forms an edge-centric TCPS. This edge-centric TCPS system can reduce data loss and data delivery delay, and fulfill the high bandwidth requirements. Within the edge-centric TCPS, Vehicle-to-X (V2X) communication, along with the in-vehicle sensors, provides a 360-degree view for CAVs that enables autonomous vehicles’ operation beyond the sensor range. The addition of wireless connectivity would improve the operational efficiency of CAVs by providing real-time roadway information, such as traffic signal phasing and timing information, downstream traffic incident alerts, and predicting future traffic queue information. In addition, temporal variation of roadway traffic can be captured by sharing Basic Safety Messages (BSMs) from each vehicle through the communication between vehicles as well as with roadside infrastructures (e.g., traffic signal, roadside unit) and traffic management centers. In the early days of CAVs, data will be collected only from a limited number of CAVs due to a low CAV penetration rate and not from other non-connected vehicles. This will result in noise in the traffic data because of low penetration rate of CAVs. This lack of data combined with the data loss rate in the wireless CAV environment makes it challenging to predict traffic behavior, which is dynamic over time. To address this challenge, it is important to develop and evaluate a machine learning technique to capture stochastic variation in traffic patterns over time. This dissertation focuses on the development and evaluation of various connected and autonomous vehicles applications in an edge-centric TCPS. It includes adaptive queue prediction, traffic data prediction, dynamic routing and Cooperative Adaptive Cruise Control (CACC) applications. An adaptive queue prediction algorithm is described in Chapter 2 for predicting real-time traffic queue status in an edge-centric TCPS. Chapter 3 presents noise reduction models to reduce the noise from the traffic data generated from the BSMs at different penetration of CAVs and evaluate the performance of the Long Short-Term Memory (LSTM) prediction model for predicting traffic data using the resulting filtered data set. The development and evaluation of a dynamic routing application in a CV environment is detailed in Chapter 4 to reduce incident recovery time and increase safety on a freeway. The development of an evaluation framework is detailed in Chapter 5 to evaluate car-following models for CACC controller design in terms of vehicle dynamics and string stability to ensure user acceptance is detailed in Chapter 5. Innovative methods presented in this dissertation were proven to be providing positive improvements in transportation mobility. These research will lead to the real-world deployment of these applications in an edge-centric TCPS as the dissertation focuses on the edge-centric TCPS deployment strategy. In addition, as multiple CAV applications as presented in this dissertation can be supported simultaneously by the same TCPS, public investments will only include infrastructure investments, such as investments in roadside infrastructure and back-end computing infrastructure. These connected and autonomous vehicle applications can potentially provide significant economic benefits compared to its cost