Article Preview
TopIntroduction
Smart cities have been an ever-growing concept that has created a knock-on effect across software research, leading into fields such as big data, embedded systems, cyberinfrastructure, and artificial intelligence. The promise of delivering a greener, safer, and more aware city infrastructure has captured the attention of researchers and city planners alike. Especially through the recent advancements in big data, the Internet of things (IoT), and artificial intelligence, the vision of smart cities is becoming increasingly real. However, the construction of a smart system governing something so impressively massive, like a city’s infrastructure comes with its own set of issues.
By now, it is no surprise that there are numerous challenges in the movement, storage, and analysis of massive, high-velocity, and abstract data. The development of these software systems often faces considerable constraints imposed by physical system restrictions such as limited broadband Internet access or available computational resources. Typically, these constraints are imposed on systems that involve deployments of edge devices, or endpoints of the larger system, for data acquisition. Also, these larger software systems are expected to provide some distributed computing environment for data processing purposes. It is because of these compounding constraints that researchers and other project personnel are forced to fine-tune and develop a sustainable and scalable data pipeline carefully.
The rest of the paper is structured as follows. This paper begins by addressing the main problems, followed by a section on its contributions. Next, it covers the background and related works relevant to the study. The following sections explain the system’s design and implementation, present the results of its data services, and discuss these findings in detail. Afterward, the conclusion reviews the study’s overall impact, and the final section outlines potential directions for future work.
Problem Statement
As the concept of smart cities continues to evolve, the need to manage complex urban infrastructure increases over time. This brings about several issues such as limited broadband, computational resources, and storage. These systems rely on the integration of tomorrow’s hardware while incorporating it into yesterday’s infrastructure. However, building the infrastructure faces challenges in the movement, storage, and data analysis.
The core issue lies in identifying a scalable data pipeline that can operate under the physical limits of the edge devices and the network. These constraints often limit the performance of distributed systems for processing data across an IoT ecosystem. These constraints must be addressed to ensure the scalability, efficiency, and effectiveness of an implementation. To achieve this, the implementers of such systems must spend considerable effort tuning different components of the infrastructure to ensure a robust and efficient system.
Contribution
In this paper, we present a multi-format point cloud data pipeline to serve the smart city infrastructure within Reno, Nevada, an extension of the functionalities from Carthen et al. (2024a, 2024b). The pipeline was designed to allow smart city researchers to build software involving pedestrian, vehicle, and object detection. In our pipeline, we implemented Google Draco and LAZ as point cloud formats with differing Lossy formats (Sato et al., 2020). A variety of formats were included in the pipeline’s design to enable multiple forms of point cloud compression and to allow for various levels of detail when considering a lossy format. A specialized suite of customized applications and web services were developed, such as a metadata service that pulls from the infrastructure, as well as several web download services for several point cloud formats. Along with these services, a GraphQL service was implemented to facilitate interfacing with the metadata, packet capture (PCAP), and ROS 2 web services at different intersections. To evaluate the effectiveness of the system’s streaming capabilities with point cloud formats, a comparison of Draco, LAZ, and PCD was also conducted.