Timely information sharing in communication constrained systems

In the near future it is envisaged that there will be a proliferation of disruptive applications combining sensing capabilities with cutting-edge wireless technologies. The wide deployment and availability of sensing nodes as well as the large amounts of data being collected will call for the design of "smarter" ways of gathering and processing such data. Such networks will be driven by the need to extract the most relevant data/information that is of interest to possibly multiple agents/nodes, while optimizing the allocation of the shared limited available resources to adapt to the heterogeneity of the interests of the different nodes as well as the heterogeneity in their network conditions. Indeed, the timely sharing of relevant information is shaping to be crucial in applications where real-time decisions are to be constantly made. In the automotive industry for example, it is expected that vehicles equipped with sensing nodes could collaborate by sharing sensed information which would allow for vehicles with obstructed views to make safer decisions and/or enhance the capacity of roadways. In the infotainment industry, and more specifically in the case of Virtual Reality (VR) applications that require large amounts of data to be constantly streamed, users at proximity of each other and that are part of the same VR experience may largely benefit from sharing resources such as edge caches that could be leveraged for the timely computation and delivery of the needed data especially in settings where different users may request the same data. The optimization of information sharing in communication constrained systems will thus be a fundamental problem underlying such systems. The focus of our work is on the modeling and analysis of these classes of problems and their implications in practical sensing systems. This thesis is composed of two main parts. In the first one, we explore the optimization of applications where timely sharing of information leads to enhanced safety and more accurate real-time decisions. We investigate novel metrics and algorithms aimed at achieving a high degree of real-time situational awareness in applications with distributed sensing nodes. In the second part of this thesis, we explore the timely sharing of information in a multi-user VR setting that would provide immersive VR experiences among the users. In particular, we explore how to support 360° VR video applications where the prediction of users' future viewings orientation is a major component as well as how to leverage overlaps in users' predictions in order to relieve the load on the shared communication network resources. Such applications face major challenges mainly tied to the heterogeneity in users' devices, predictions, network conditions, etc., which we propose to tackle through the careful design of metrics and policies robust to such heterogeneity and aimed at enhancing while achieving fair VR performance among the users.