Research & engineering

Research

Communication networks, IoT and cyber-physical systems, learning and control, smart infrastructure, and wireless sensing — organized into five connected research areas.

Fig. 01 · IAC 2026 Figure 1 from “Goal-Oriented Bundle Management and Flow Control for Deep-Space Communications via Reinforcement Learning”: Earth–Relay–Moon–Mars network
Figure 1 reproduced from S. Baghaee et al., “Goal-Oriented Bundle Management and Flow Control for Deep-Space Communications via Reinforcement Learning,” 77th International Astronautical Congress (IAC 2026), 2026. View paper.
01

Communication & networking

Protocol and flow-control design for links that are slow, shared or intermittent. The thread runs from measuring Age of Information on real TCP/IP and LoRa links, through application-layer forward error correction that holds peak age below a threshold, to goal-oriented transmission and freshness-driven satellite IoT access. Current work in the ERC-funded GO-SPACE project extends this to delay-tolerant and deep-space settings, where feedback is slow and contact windows are scheduled.

16 recorded works2018–2026

Latest: goal-oriented bundle management (IAC 2026) GO-SPACE — goal-oriented flow control A³L-FEC — age-aware application-layer FEC AoI in future networks (Huawei)

Age of InformationProtocol DesignForward Error CorrectionSatellite Communications

Representative works

Fig. 01 · SIU 2025 Figure 1 from “Ontology-Driven Smart Building Semantics: An EnergyPlus-SAREF Approach”: The pipeline for transforming EnergyPlus data into a SAREF-based knowledge graph, illustrating the key steps from raw data extraction to semantic querying
Figure 1 reproduced from U. Kırnapcı et al., “Ontology-Driven Smart Building Semantics: An EnergyPlus-SAREF Approach,” 2025 33rd Signal Processing and Communications Applications Conference (SIU), 2025. View paper.
02

IoT & cyber-physical systems

End-to-end IoT work: embedded nodes and firmware, LoRa/LoRaWAN and MQTT pipelines, gateways and edge processing, and the data platforms above them. Recent work covers environmental sensing platforms deployed in buildings, ontology-based smart-building semantics that make sensor data portable between tools, and digital twins that keep a model in step with the physical system it describes.

19 recorded works2018–2026

SUIT SensorBox & digital twins MipMap digital twin platform LoRa IoT deployments at JeoIT

IoTDigital TwinSmart BuildingsLPWAN / LoRa

Representative works

Fig. 01 · SIU 2025 Figure 1 from “Analysis of Model-Agnostic Meta-Reinforcement Learning on Automated HVAC Control”: MAML-RL framework for HVAC automation
Figure 1 reproduced from U. Filiz et al., “Analysis of Model-Agnostic Meta-Reinforcement Learning on Automated HVAC Control,” 2025 33rd Signal Processing and Communications Applications Conference (SIU), 2025. View paper.
03

AI, learning & control

Learning used where a decision has to be made under constraints. Reinforcement learning for building climate control, meta-reinforcement learning for transferring a policy between climates rather than retraining from scratch, LSTM surrogate simulators that shorten the training loop, and edge inference placed on the device so that only the conclusion has to travel.

11 recorded works2018–2026

RL-based deep-space flow control Learning-based HVAC control Building Smart Energy Management System GridAI — edge-AI fault detection

Reinforcement LearningHVAC ControlMachine LearningEdge AI

Representative works

Fig. 01 · GridAI 2026 Figure 1 from “GridAI: Edge AI Smart Grid Fault Detection with Goal-Oriented Communication”: End-to-end overview of the GridAI system
Figure 1 reproduced from M. B. Aydar et al., “GridAI: Edge AI Smart Grid Fault Detection with Goal-Oriented Communication,” Proceedings of the 24th Annual International Conference on Mobile Systems, 2026. View paper.
04

Smart infrastructure & energy

Instrumenting infrastructure that cannot easily be serviced. An AoI-aware fault-indicator device for electricity distribution, taken from hardware and embedded software through to pilot deployment in transformer substations, and an edge-AI detector that decides locally what is worth transmitting. Behind it sits a decade of energy-harvesting and wireless sensor network work on running nodes without batteries.

12 recorded works2012–2026

Age-sensitive grid monitoring (FreshData) E-CROPS — energy-harvesting networks Energy-efficient design of wireless networks

Smart GridEnergy HarvestingWireless Sensor Networks

Representative works

Fig. 01 · SIU 2026 Figure 1 from “Evaluating the Impact of CSI Preprocessing on WiFi-Based Human Activity Recognition”: System design and setup for CSI data acquisition
Figure 1 reproduced from Y. C. Çelik et al., “Evaluating the Impact of CSI Preprocessing on WiFi-Based Human Activity Recognition,” 2026 34th Signal Processing and Communications Applications Conference (SIU), 2026. View paper.
05

Wireless sensing

Reading the physical world from signals that are already there. Wi-Fi channel state information for device-free human activity recognition, including a synchronised acquisition pipeline, a public dataset and a controlled architecture benchmark; and earlier work on magnetic sensor networks for detecting, tracking and identifying ferromagnetic targets, alongside capacitive and dielectric material characterisation.

7 recorded works2012–2026

Wi-Fi CSI sensing and activity recognition Magnetic wireless sensor networks Material characterisation for recycling

Wireless SensingHuman Activity RecognitionMagnetic SensingTarget Localization

Representative works

Areas overlap by design: a smart-grid monitor is an IoT deployment, an edge-AI system and a networking problem at once. The counts above come straight from the topics recorded against each work, so a paper is counted in every area it genuinely belongs to.