Integrated smart pavement systems for environment monitoring, localization, and traffic data collection

Abstract

The traffic engineering is making considerable strides towards autonomous driving technology. Although single-vehicle autonomous driving (SAD) has advanced algorithms and sensing technology, it still has limitations on reliability and robustness, particularly in challenging weather conditions and high-density traffic. In contrast, vehicle-infrastructure cooperative autonomous driving (VICAD) can overcome many of these challenges and the key to development of VICAD is to build high-level smart roads. To address this challenge, this thesis proposes an Integrated Smart Pavement (ISP) system. Specifically, this thesis begins with the overall framework of the ISP system, including lab and practice demonstrations that assess the system’s feasibility. Following the introduction of the overall framework, this thesis proceeds to describe in detail the various incorporated techniques of the proposed ISP system. Firstly, an artificial-intelligence-of-things (AIoT)-based framework with ISP for water quality monitoring and estimation of crucial parameters is proposed. The chapter conducts an extensive literature review to identify critical water quality parameters that inform the development of a data-driven framework. The framework includes two artificial intelligence models used to estimate unmeasurable water quality parameters and construct a real-time water quality monitoring AIoT system. A case study is conducted to validate the proposed framework, and the overall system performance is satisfactory. The prediction performance is better at river sites with higher pollutant levels. Afterwards, this thesis proposes an ISP-embedded novel pavement marking for autonomous vehicle localization and describes it in technical detail. The marking is customized, a detector is trained using YOLOv5-based object detection, and a decoding algorithm is proposed that processes marking information in real-time. The study conducted two sets of road trials to validate the efficacy of the novel marking and associated algorithms. The system’s performance was evaluated under different environmental conditions, and it was found that depression angle is the critical factor affecting performance. In the next part, this thesis focuses on exploring the potential of ISP-embedded IRS in future 6G systems for delivering ubiquitous and reliable communication and accurate localization information. This part of study proposes a deep reinforcement learning (DRL) based framework to optimize the IRS configuration selection problem in localization process and presents a protocol for implementation. Simulation results indicate that increasing the number of IRS configurations can reduce localization errors, and with the proposed DRL-based methodology, more than a 40% improvement can be achieved. The proposed methodology exploits the localization performance in future 6G systems for intelligent transportation systems (ITS). Finally, this thesis proposes an ISP-edge-deployment solution for real-time drones-assisted turning movement counts (TMC) collection using the YOLOv5-StrongSORT-TMC algorithm in response to the increasing demand for traffic data collection. A case study is conducted at a busy intersection in Hong Kong to evaluate the method’s effectiveness. The results outperform similar approaches presented in previous studies. In summary, the proposed ISP system offers a solution to the challenges facing conventional transportation infrastructure. By implementing adaptive traffic systems and embracing advanced technologies, the ISP system improves transportation infrastructure’s capabilities and enhances its relationship with the surrounding environment.