Prof. Roshan G. Ragel supervises postgraduate research students pursuing PhD, MPhil, and MSc degrees at the University of Peradeniya and in collaboration with other institutions, as well as undergraduate students in their final year research projects. Research areas span Internet of Things, Artificial Intelligence & Machine Learning, Bioinformatics, Computer Vision, Wearable Computing, and Engineering Education.
Postgraduate Students
PhD Candidates
Mathuranthagaa Selvaratnam
PhD (Full Time)
Co-supervised with Muttukrishnan Rajarajan, City, University of London
Romiyal George
PhD (Full Time)
Plant Diseases Detection
Co-supervised with Thuseethan Selvarajah
M Santhanakrishnan
PhD (Full Time)
Co-supervised with Thuseethan Selvarajah and Siyamalan Manivannan
Kanagasabai Thiruthanigesan
PhD (Full Time)
Roshan Pushpakumara
PhD (Part Time)
Co-supervised with Isuru Nawinne
S Nimishan
PhD (Full Time)
Co-supervised with Thuseethan Selvarajah and S Vasanthapriyan
Niroshan Herath
PhD (Part Time)
Started: July 22, 2024
Co-supervised with Isuru Nawinne
Nadeeshani Aththanagoda
PhD (Full Time)
Proposal submitted
Co-supervised with Dr. Kulathilaka (RJT) and Sahah (UCL)
Lohara Chathumini
PhD (Full Time)
Explainable Anomaly Detection in Text Data Using Large Language Models in Information Systems
Co-supervised with Thuseethan Selvarajah
R Sakuntharaj
PhD (Part Time)
Venuja Niranjan
PhD (Full Time)
Co-supervised with Prof Siyamalan
Suhail Razeeth
PhD (Full Time)
Co-supervised with Dr. S. Tuseethan and Dr. M.M.M. Mufassirin
Thilini Bakmeedeniya
PhD (Full Time)
Co-supervised with Dr Selvarajah Thuseethan
Sotheeswaran
PhD (Full Time)
MPhil Candidates
Rukshani Puvanendran
MPhil (Full Time)
Co-supervised with Asanka P. Sayakkara (UCSC)
Rasendram Muralitharan
MPhil (Full Time)
Co-supervised with Upul Jayasinghe
Sandaruwan Herath
MPhil (Part Time)
Co-supervised with Prabhath Gunathilake
Warunika Hippola
MPhil (Part Time)
Co-supervised with Isuru Nawinne
Hasini Vindiya Athauda
MPhil (Part Time)
Co-supervised with Upul Jayasinghe
H Hashan
MPhil (Full Time)
Co-supervised with Asitha Bandaranayake
Dinusha Premasiri
MPhil (Part Time)
Co-supervised with Isuru Nawinne
Balenthira Thanusanth
MPhil (Full Time)
Approved on May 28, 2025 at HDC meeting
Started: December 12, 2024
Enhancing Explainability and Privacy in Medical Image Analysis using Federated Learning and Explainable AI
Co-supervised with Thuseethan Selvarajah, Ayesh, and Bimali
Chamika Dissanayake
MPhil (Part Time)
Proposal submitted
Co-supervised with Dr Kartheeswaran
Akila Udara
MPhil (Part Time)
Submitted to PGIS
Co-supervised with Prof Saku
Sanjaya Wijerathna
MPhil (Full Time)
Under Discussion
Co-supervised with Siyamalan and Kokul
Dinuka Mudalige
MPhil (Full Time)
Under Discussion - Registering at OUSL
Integration of AI and IoT techniques for Resource Optimization in Aeroponic Orchid Cultivation
Co-supervised with Dr. S. N. C. M. Dias and Dr. D. D. M. Ranasinghe
Hansani Umayangana
MPhil (Part Time)
Under Discussion
Niroji Thayalan
MPhil (Full Time)
Approved on May 28, 2025 at HDC meeting
Started: March 18, 2025
Co-supervised with Dr Selvarajah Thuseethan
Vithursan T
MPhil (Full Time)
Approved on May 28, 2025 at HDC meeting
Co-supervised with Prof Priyan and Dr Thuseethan
Thiksiga Ragulakaran
MPhil (Part Time)
Co-supervised with Roshan Ragel, PE, and VH
Alan Amath
MPhil
Provisional registration
Co-supervised with Dr. S. Tuseethan and Dr. M.M.M. Mufassirin
MSc Candidates
Nelofar Akaliya
MSc (Part Time)
Co-supervised with Bimali Weerakoon
U Ahamed
MSc
Registered
Identify the cyber security-based challenges on librarianship at the implementation of digital library services in Sri Lankan Universities
Co-supervised with Dr P Wanigasuriya (Dep Lib & Info Science, KLN)
Undergraduate Students
E20 Research Projects
Reducing Context Switching Latency in RTES
Using register file partitioning to eliminate save/restore overhead for high-priority tasks in RISC-V embedded systems.
View Project
Synthetic Cardiac MRI Image Generation using Deep Generative Models
Generating anatomically accurate synthetic cardiac MRI images using GANs, VAEs, and diffusion models to address scarce annotated medical data.
View Project
E19 Research Projects
A Behavioural Plugin Tool For Fake News Identification
Multimodal browser extension using Hybrid-BERT to detect fake news in real-time via emotion, sentiment, and geographic analysis of Reddit posts.
View Project
AI Driven Latency Constrained Resource Management In Kubernetes
Reducing Kubernetes resource over-provisioning for microservices while keeping latency within SLOs using RL, Bayesian optimization, and critical reduction point analysis.
View Project
AI Driven Decision Making Models for Corporate Applications
GenAI model trained on organizational data to automate HR tasks — document generation, automated emails, and employee guidance using LangGraph and GPT-4o.
View Project
AI Powered Knowledge Management System
RAG-based knowledge management system for NRENs using vector embeddings, LangChain, and GPT to manage institutional knowledge and training materials.
View Project
Classification of Motor Imagery EEG based Tasks for Brain Computer Interfacing Applications
Single-joint (wrist) motor imagery EEG classification using CNNs, Transformers, and hybrid architectures for improved BCI precision in neurorehabilitation.
View Project
Enhancing Multimodal Fusion Techniques for Depression Detection
Web platform with AI chatbot for depression screening, handling missing modalities and class imbalance using audio, video, and text multimodal fusion.
View Project
LowComplexity Algorithms For EnergyEfficient Arrhythmia Classification In Wearable Devices
Low-complexity ECG arrhythmia detection algorithm for resource-constrained wearables with channel attention modules for improved classification accuracy.
View Project
Peraswarm Active Localization and Navigation
Multi-robot C-SLAM framework combining EKF sensor fusion, RTAB-Map, A* global planning and DQN local collision avoidance for UGV/UAV teams in GPS-denied environments.
View Project
Putting Privacy into Practice
AI-powered GDPR-compliant serious game platform with adaptive RL difficulty and LLM feedback to train software developers in privacy-aware coding.
View Project
Quality Assessment of Final Root Canal Treatment Using Intra Oral Periapical Radiographs
Deep learning tool using BiomedCLIP and MedSAM to automatically detect and evaluate root canal treatment quality from IOPA radiographs.
View Project
RISC-V Based ASIP for Accelerating DNA Sequence Alignment on Embedded Systems
Custom RISC-V co-processor via RoCC interface targeting Minimap2's chaining stage for energy-efficient third-generation DNA sequence alignment on embedded systems.
View Project
Solve Issues In Large Code Repositories
Novel SWE-Bench methodology combining graph-based repository representation, RAG with Stack Overflow, and deep LLM iterative reasoning for cost-effective issue resolution.
View Project
Verilog Code Generation With Variations Using Fine-Tuned Large Language Models
Automating Verilog test case variation generation via fine-tuned LLMs using Python-based transformations, direct Verilog transformations, and X-Form functions.
View Project
E18 Research Projects
An Empirical Study On Prompt Injection Attacks And Defences
Empirical analysis of prompt injection attack vectors against LLMs and evaluation of defence mechanisms to improve robustness and security.
View Project
Automated Medical Image Annotation for Dataset Building
Comparative study on generalizable few-shot segmentation models (SAM, SegGPT, UniverSeg) for automated oral and medical image annotation without retraining.
View Project
DL-Based Stereo Matching Estimation
Deep learning approach for accurate stereo disparity estimation, enabling depth perception from binocular camera systems.
View Project
Extending and Implementing Process Mining Techniques – Prevention and Mitigation
PraeclarusPDQ plug-in using the Odigos framework to identify root causes of data quality issues in event logs and generate mitigation strategies.
View Project
GPU Acceleration for Deep Learning based Comprehensive ECG Analysis
1D-CNN model using transfer learning from synthetic DeepFake ECG data to real PTB-XL signals for GPU-accelerated cardiac disease classification.
View Project
Generating Patient Clinical Letters Using LLMs
Training and evaluating large language models for automated generation of accurate, clear, and privacy-preserving patient clinical letters.
View Project
Machine Vision For Quality Inspection
Real-time computer vision system for automated toy inspection using depth cameras — detecting defects in eye paint, alignment, and gaze direction on assembly lines.
View Project
Neuromorphic NoC Architecture for SNNs
Scalable RISC-V based Network-on-Chip for spiking neural networks implemented on FPGA, optimized for low-power neuromorphic computing.
View Project
PeraSwarm: Simultaneous Localization and Mapping in Mixed Reality Environment
Swarm robotics framework combining physical and virtual robots in a mixed reality environment for collaborative SLAM with Voronoi coverage and occupancy grid mapping.
View Project
Using AI in Determining Alveolar Bone Loss in Dental X-rays to Aid in Periodontal Diagnosis
AI-assisted tool using object detection and segmentation to automatically assess alveolar bone loss severity and patterns in IOPA dental radiographs.
View Project
Virtual Patient Simulator for Skill Training in Dentistry
Enhanced web-based VPS with custom case creation for dental tutors, semi-automated Unity/WebGL 3D models, and dental chart marking evaluated via HCI usability testing.
View Project
E17 Research Projects
Acceleration of Deep Neural Network Based Nanopore Super Accuracy Basecalling
Accelerating the Slorado nanopore basecaller via contiguous RNN weight memory optimization achieving ~52% speedup, with investigation of Winograd-based tiled matrix multiplication.
View Project
Comprehensive ECG Analysis with Deep Learning on GPU Accelerators
AICRN deep learning architecture using spatial and channel attention for automated ECG parameter regression (PR, QT, QRS, HR) on GPU as an open-source MUSE alternative.
View Project
Improving Trajectory Guided Control Prediction for Autonomous Driving
Identifying and improving accuracy issues in the TCP architecture for autonomous driving using the CARLA simulator.
View Project
Large Language Models in Education
Exploring LLMs to develop an intelligent tutoring system that personalises learning and improves student engagement in higher education.
View Project
Low Complexity Algorithm for Arrhythmia Detection
Low-complexity, low-memory QRS complex and arrhythmia detection algorithms for microcontrollers with severely constrained resources (<2kB SRAM, 1–8 MHz).
View Project
Music Emotion Recognition
Automatic analysis and classification of emotional content in music using song-level MER and music emotion variation detection (MEVD) with deep learning models.
View Project
Neuromorphic NoC Architecture for SNNs
Novel RISC-V based Network-on-Chip architecture for simulating spiking neural networks in a scalable, event-driven manner, implemented and validated on FPGA.
View Project
OCR and Translation from Sinhala to Tamil for Printed Documents
Web application using Tesseract OCR and Google Translate API to convert printed Sinhala documents (images, PDFs, text) into editable Tamil text with 85% OCR and 77% translation accuracy.
View Project
S2TPVFormer: Spatiotemporal 3D Semantic Occupancy Prediction
Spatiotemporal transformer architecture (S2TPVFormer) with novel Temporal Cross-View Hybrid Attention for 3D semantic occupancy prediction, achieving +4.1% mIoU gain on nuScenes (AAAI Workshop 2025).
View Project
Virtual Patient Simulator for Skill Training in Dentistry
Unity-based 3D virtual dental patient simulator enabling students to practice patient assessments, outside/inside mouth examinations, dental tools and X-ray interpretation with grading systems.
View Project
Programming and Compiler Toolchain for Multi-Agent Systems
Integrated Development Environment for swarm robots with block-based visual programming (Google Blockly), dynamic C++/Java code generation, and OTA deployment via MQTT for both virtual and physical platforms.
View Project
Network Anomaly Detection
Structure learning in Graph Attention Networks (GATs) with causal-guided adaptive attention for improved network anomaly detection while reducing false positives.
View Project
Novel Compression Methods for Nanopore Raw Signal Data
Exploring lossy compression approaches for nanopore raw signal data by identifying signal noise and its impact on downstream basecalling, targeting reduced long-term archival costs.
View Project
E16 Research Projects
Acceleration of DTW Algorithm for Real-Time Nanopore Selective Sequencing Using GPUs
Optimizing the Dynamic Time Warping (DTW) algorithm and accelerating it using GPUs for real-time nanopore selective sequencing on GPU-equipped laptops or embedded devices like NVIDIA Jetson.
View Project
Enhancing Video Game Experience With Brain Computer Interface
EEG-based BCI using inner speech paradigm to decode four directional commands (Left, Right, Up, Down) for controlling a navigation video game, using ICA, CWT and CNN transfer learning on ResNet50.
View ProjectHardware Cache Switching with Operating System Context Switches
Cache partitioning technique dividing the cache into runtime-selectable partitions to reduce inter-process interference and minimise context switch overhead in embedded systems.
View Project
Human/Animal Emotion Detection Using ECG and ML Techniques
2D CNN models trained on a custom 9-emotion ECG dataset from 25 subjects achieving 70% valence and 65% arousal accuracy, with extended experiments detecting dog emotions using the same model.
View Project
Smart Agriculture: Use of Computer Vision, ML and IoT to Improve Crop Productivity
Smart greenhouse system integrating IoT, computer vision and ML for plant disease and anomaly detection to enable site-specific interventions and maximise crop yield.
View Project
Virtual Patient Simulator for Skill Training in Dentistry
Unity and Blender-based 3D virtual dental patient simulator where students practice outside/inside mouth examinations, use digital dental tools and interpret X-rays in a safe, low-cost environment.
View Project
Constructing Quantum-Safe Leakage-Resilient Public Key Encryption Scheme
Quantum-secure public-key encryption based on lattice hash proof systems that remains secure even under partial key leakage from side-channel attacks or poor key management.
View Project
Neuromorphic Architecture
Brain-inspired computer architecture with multiple RISC-V processors interconnected like neurons via a scalable event-driven NoC, optimised for low-power neuromorphic computations such as character recognition.
View Project
Open World Object Detection and Discovery
Tackling open-world object detection by simultaneously detecting known categories and identifying unknown objects, with incremental learning of new classes as labelled data progressively becomes available.
View Project
Post-Quantum Cryptographic Schemes Based on Plain Lattices
Cryptographic scheme using plain lattices with the Learning With Errors problem for quantum-resistant encryption on classical computers, implementing lattice trapdoor functions and proposing efficient key sizes.
View Project
Swarm Intelligence Programming Framework
High-level programming and compiler framework for the PeraSwarm project enabling algorithm creation via "writing" or "drawing" with automatic translation to low-level code for multiple real and virtual robot hardware types.
View Project
E15 Research Projects
Accelerating Adaptive Banded Event Alignment Algorithm with OpenCL on FPGA
OpenCL-based FPGA implementations of the ABEA algorithm for nanopore DNA sequencing, achieving better energy efficiency than GPU counterparts while maintaining competitive runtime performance.
View Project
Brain Computer Interface for Controlling Virtual Objects Using Self-Paced Mind Intent
Self-paced EEG-based BCI system analysing non-motor imagery mental tasks using ML pipelines to control virtual objects in a Unity environment in real-time.
View Project
Improved Gesture Recognition for sEMG Based Human Machine Interface
sEMG-based hand gesture recognition using time, frequency and time-frequency domain features with ML classifiers (SVM, RF, KNN, RNN), achieving 90.03% accuracy for 8-gesture classification on a 24-channel dataset.
View Project
Mixed Reality Based Simulation Platform for Swarm Robotics
Mixed-reality simulator allowing physical and virtual robots to collaboratively run swarm intelligence algorithms in real-time, combining decentralised autonomous control with biological-inspiration principles.
View Project
Real-Time Emotion Recognition Using Electrocardiogram Analysis
CNN trained on the DREAMER ECG dataset to predict valence, arousal and dominance values in real-time, mapping them to a 3D emotional plane, with preliminary analysis extended to animal emotion recognition.
View Project
Revealing miRNA Biomarkers for Alzheimer's Disease Using NGS
Identifying differentially expressed miRNA biomarkers for Alzheimer's Disease from the NCBI GSE46579 NGS dataset using Galaxy preprocessing, statistical analysis (p-value, fold-change, AUC) and ML-based feature selection.
View Project
Near-IR Spectroscopy for Vein Visualization
Low-cost near-infrared vein visualisation device using NIR LEDs, a modified CMOS camera and OpenCV (CLAHE, histogram equalisation) displaying real-time vein maps on a 7-inch IPS LCD, optimised for Asian skin tones.
View Project
