The proliferation of cost-effective sensors, coupled with the rapid advancement of networking technology, has spurred the development of numerous real-time networked applications. These include disaster response, aerial surveillance, smart grids, and building automation. In these applications, data from multiple sources are transmitted across networks, which may be one-hop or multi-hop, and exhibit static or dynamic topologies. Ensuring the timely delivery of data from all sources is critical for real-time monitoring and decision-making, thereby enabling the information holder to maintain up-to-date information. This dissertation explores efficient transmission and scheduling schemes for two types of networks to enhance network and information freshness performance.

Firstly, the network with one-hop static topology is considered, where multiple information sources compete for a shared channel to send their data packets to the base station (BS). Due to the limited transmission resources, the BS has to judiciously schedule the transmission from each source. However, this presents the following challenges: 1) The channel can be unreliable in wireless environments, i.e., the data packet transmitted over the channel might drop with some probability. Furthermore, the probability can be unknown to the BS and different across sources. As packet loss hinders the BS from updating its held information for a source, it has to jointly learn and adapt to the channel unreliability during the transmission scheduling; 2) Most existing information freshness metrics assume the data packets from a source are equally significant. In contrast, for systems like the Artificial Intelligence of Things (AIoT), raw data is analyzed by intelligent algorithms deployed near the information sources to generate event summaries. Consequently, the conventional freshness metrics are inadequate to characterize the timeliness of capturing events whose occurrence rate might be time-varying. This necessitates the development of a new freshness metric and corresponding transmission scheduling algorithms. To address the former challenge, we first derive the optimal stationary randomized sampling algorithm that maximizes the throughput while satisfying per-source information freshness constraint given the channel unreliability, and then develop learning algorithms with guaranteed feasibility and bounded throughput regret (gap between performance under the learning algorithm and the oracle algorithm) without knowing the channel unreliability. Furthermore, we also propose a learning algorithm with guaranteed regret and feasibility that can adapt to unsatisfiable information freshness constraints. To address the latter challenge, we first propose a new information freshness metric named age of event (AoE) to characterize the timeliness of capturing events, and then design scheduling algorithms when the knowledge of event occurrences is fully known, partially known, and completely unknown with theoretical performance analysis.

Secondly, a typical network with multi-hop dynamic topology, the UAV network, is considered. Due to the high mobility of UAVs, the physical network topology varies frequently. Consequently, the mainstream transmission scheme adopts a dynamic routing protocol to find the route for sending packets from the source to the destination. However, the route is not guaranteed to be viable upon the forwarding. This can lead to a high packet loss ratio and high latency, resulting in deteriorating information freshness. As a remedy, we propose a UAV topology control scheme to maintain a fixed communication topology during the UAV movement to enable static routing. This scheme includes 1) a distributed motion planning algorithm that optimizes the coverage ability of the UAV swarm subject to the constraint of keeping the given communication topology intact, and 2) a topology management protocol to allow reconfiguration of the communication topology during the UAV movement. Simulation results demonstrate both the packet loss ratio and the latency are significantly improved under our solution, and the UAV swarm also exhibits an efficient coverage ability.

The approaches proposed in this dissertation contribute to both the fields of networking and real-time systems. The transmission scheduling algorithms guarantee the information freshness of each source in the wireless environment, and ensure the timely capture of event information even when the event occurrences are unknown. The UAV topology control scheme fundamentally improves the performance of UAV networks by enabling static routing within dynamic physical topology.