This proposal aims at a joint consideration of information generation, processing, transmission, and reconstruction in beyond 5G (B5G) massive and mission critical IoT networks by taking into account the age (AoI) and value of information (VoI). So far, in 5G networks, those processes are treated separately, and the importance and usefulness of the generated and transmitted information are ignored. In this project will develop theoretical and algorithmic foundations of goal-oriented, data importance-aware communication to depart from the separated and conventional content-agnostic paradigm which will help to reveal the potential of future hyperconnected intelligent systems.
The main research challenge is how to efficiently operate a massive IoT network, particularly in a mission critical setup. We will investigate how to get the exact information that is needed where it needs to be, when it needs to be there, without the tremendous overheads of current systems focused on simply delivering data, with no regard to its value to the receiving application.
The PhD student will design concrete metrics of assessing information significance. Furthermore, we will develop multiple access and transmissions schemes. The development of optimal end-to-end information handling schemes will involve joint information generation, transmission, and reconstruction of processes. The student is expected to have strong mathematical background in particular on stochastic modelling, optimization, and reinforcement learning. Prior knowledge on Age and Value of Information will be considered as a big plus. This PhD project consists of the following work packages.
We will establish metrics for assessing information significance, such as the AoI/VoI-based metrics. These metrics will jointly consider information sampling and transmission, which are relevant to smart IoT scenarios. Developing an AoI to be transmission-aware provides a generalization that departs from the classical evolution of AoI over time, furthermore, taking into account the evolution of the source goes toward a timing-value aware metric.
In this part, we will focus on how multiple devices can access a shared medium optimally. When information freshness is important, we know that the policy of generating traffic at will can be optimal under specific conditions. Thus, we do not have to retransmit the packets that fail to be received, since a newer sample will be acquired. In that case, sampling (information generation) must be considered jointly with the transmission, which differentiates our setting from the classical multiple access schemes such as grant free massive access. Importance-aware access, where the devices can adapt their transmissions based on the novelty of the acquired sample is another important direction. This will result to fewer transmissions without sacrificing the reconstructed accuracy or violating the accuracy requirements when real-time monitoring and reconstruction is the purpose of communications.
In this part, we will develop optimal end-to-end information handling schemes, which will enable joint information generation, transmission, and reconstruction of multimodal information. Consider a scenario where multiple sensors collect similar information; by utilizing smart algorithms, we can prefilter and collect data only from a small subset of the sensors that when they will be activated, will sustain the required threshold of AoI/VoI. Also, taking into account the heterogeneity among sensors, the collected data can have different quality in terms of precision or freshness. Consideration of sampling costs in terms of energy or availability of the observing source is another important aspect to take into account and it will affect the scheduling and resource management schemes. Consider for example the case where you can sample from a sensor that has fresh information but of low quality, or you can retrieve less fresh information but of higher quality. Clearly, the decision of activation will depend on the target we want to achieve.