Research Projects
Click any project to expand details. Funded projects and active research initiatives.
Active Projects
2025 – PresentThis project focuses on building a Brain Computer Interface (BCI) system that can assist severe mobility-impaired persons by enabling wheelchair control through neural signal interpretation. The core objective is to develop a safe, reliable, and adaptive assistive platform that improves independence and quality of life.
Key research components include EEG signal acquisition, preprocessing, real-time intent decoding using machine learning models, and embedded system integration for wheelchair actuation. The system targets low-latency response and practical deployment in real mobility-support scenarios.
Building retrieval-augmented generation (RAG) workflows that combine structured knowledge graphs with semantic retrieval to improve factual grounding, multi-hop reasoning, and answer traceability in large language model applications.
The project investigates how graph-structured knowledge can enhance the reliability of LLM outputs by providing structured context, enabling traceable reasoning paths, and reducing hallucination in domain-specific question answering systems.
Designing graph-based deep learning models to capture spatial and temporal dependencies in brain-signal data for robust emotion-state recognition and interpretable affective computing. The project bridges neuroscience and AI by treating EEG channel correlations as graph structures.
Past & EU-Funded Projects
BRAIN-IoT focuses on complex scenarios where actuation and control are cooperatively supported by populations of IoT systems. The breakthrough targeted is a framework and methodology supporting smart cooperative behaviour in fully decentralized, composable and dynamic federations of heterogeneous IoT platforms.
BRAIN-IoT tackled business-critical and privacy-sensitive IoT scenarios subject to strict dependability requirements. It enables smart autonomous behaviour in IoT scenarios involving heterogeneous sensors and actuators autonomously cooperating in complex, dynamic tasks. Open semantic models are used to enforce interoperable operations, supported by model-based development tools. The viability of proposed approaches was demonstrated in Service Robotics and Critical Infrastructure Management use cases.
MONSOON focuses on the optimization of data-intensive processes in manufacturing and process industries, characterized by high complexity and variability. The project developed a model-based control framework that optimizes process performance while ensuring stability and robustness.
The framework is based on advanced modeling techniques including machine learning and system identification, designed to handle large volumes of data generated by industrial processes. A set of tools and algorithms for real-time monitoring and control was developed and validated through industrial case studies.
In the Linked Open Data (LOD) cloud, numerous Knowledge Bases (KBs) are shared and used for data analytics, question answering, and other tasks. These KBs evolve, and unrestrained evolution may cause data to suffer from quality issues at both semantic (contradiction) and pragmatic (ambiguity, inaccuracy) levels.
This Ph.D. research developed a data quality assessment methodology for large-scale knowledge bases. The primary emphasis was on automated quality assessment using evolution analysis — exploring (i) quality assessment using KB evolution analysis, and (ii) validation using machine learning models. The work involved collaborations with ISMB (Italy), Ontology Engineering Group at UPM (Spain), and Politecnico di Torino.