Over the last years, edge computing has become a topic of significant interest in academic and industrial research, driven by the fundamental improvements in latency, bandwidth savings and energy efficiency it offers with respect to offloading application workloads. While many envisioned applications in edge computing are relevant in the context of real-time analytics and feedback, most of research today is approaching edge computing from a best effort perspective, focusing on expected system behavior. Nevertheless, in order to make edge computing appealing for industrial applications, trustworthiness must be considered as a key design objective. Trustworthiness includes not only security, but also predictability, and safety of the edge infrastructure and of the associated applications running on top of the infrastructure.
In this context, we work on:
- Theoretical modeling of integrated communication and compute paths, with an emphasis on reliability and energy efficiency
- New applications enabled by edge computing, mostly incorporating real-time feedback
- Experimental prototyping and benchmarking of edge computing architecture.
- Check out the TECoSA website, where most of this research is done at KTH!
Critical Machine-to-Machine Communcations
For a couple years, there is more and more interest in low-latency, ultra-reliable wireless networks. This interest is triggered by new, anticipated application domains of such networks, most importantly in the areas of industrial automation, car-to-x communications, as well as new consumer electronics gadgets like virtual reality glasses. If we require latencies in the range of 1 ms or below, from a theoretical as well as from a practical perspective many new research questions arrise. Over the last five years, I have been working in this area, focusing on aspects such as cooperative schemes for low-latency systems, corresponding MAC protocols, verification and validation of low-latency protocol stacks as well as experimental prototyping of corresponding systems. Among others, this activity has lead to the design and experimental prototyping of the EchoRing system.
Currently, I am focusing most on:
- Theoretical modeling of low-latency systems via finite block-length capacity models
- Wireless /wired low-latency communication paths with in-network processing
- Centralized network architectures for sub-millisecond ultra-reliable wireless networking
- Abstractions of control systems for cross-layer optimization in low-latency systems
Stochastic Network Calculus and Age of Information
Link-layer performance evaluation of wireless systems is typically best done if it considers also queuing effects. There are several ways how to do this, among which network calculus is a fairly powerful one (but arguably also a quite involved one). Over the last ten years, stochastic network calculus approaches have been developed where the cumulative arrival and service processes of a queuing systems are captured for example by the moment-generating function or by the Mellin transform. Today, we are in a good position to extend these results to specific service processes encountered in wireless systems. More recently, age of information performance models of networks have become very fashionable, as they provide an integrated model for data generation and communication towards the freshness of the conveyed data.
Currently, in the context of network calculus and age of information I am mostly interested in:
- Interference-limited communication systems
- Wireless communication under the finite block-length regime
- Secrecy channels
- Edge-like communication systems