Contents I. Introduction
Internet of Things (IoT) technologies have been widely adopted for enhanced visibility, improved efficiency, lower cost, and remote operations in many fields, e.g., infrastructure monitoring, manufacturing and logistics systems, and disaster management [1], [2], [3]. Various IoT devices have been developed that can accurately sense the environment and automatically transmit sensing data to back-end systems using wired/wireless communications. In practice, to meet the timeliness requirements and avoid overriding historical sensing data in the on-board storage of IoT devices, the sensing data should be collected in a timely manner. Despite this is always realizable in production shop floors, commercial buildings, and urban infrastructure networks, it is a formidable challenge in scenarios such as transportation infrastructure monitoring [4], [5], [6]. Specifically, this work is motivated by the monitoring of highways, bridges, and high-speed railways, where numerous IoT devices are widely distributed in large, extreme, and wild areas with limited network coverage. Besides, the monitoring video collection issue that arises in wild animal monitoring further amplifies the problems of IoT data collection, where the cameras are always deployed in the virgin forest and national parks without network coverage and are difficult to reach by humans. In these scenarios, it would be economically impossible to build wired networks or base stations that connect all these devices to collect real-time sensing data. Even with access to mobile communication networks, the continuous wireless connections between IoT devices and back-end systems are difficult as they require each IoT device to have a sufficient power supply or frequent swapping of batteries, which poses extra high-maintenance costs, especially for those being sparsely deployed in extreme and wild environments.
