International Conference on Advancements in Computing [ICAC]
Permanent URI for this communityhttps://rda.sliit.lk/handle/123456789/312
The International Conference on Advancements in Computing (ICAC) is organized by the Faculty of Computing of the Sri Lanka Institute of Information Technology (SLIIT) as an open forum for academics along with industry professionals to present the latest findings and research output and practical deployments in computing.
The primary objective of ICAC is to promote innovative research that addresses real-world challenges and contributes to the social well-being of communities. The conference provides a dynamic platform for researchers from around the world to present groundbreaking findings, exchange ideas, and establish meaningful collaborations.
Browse
250 results
Filters
Advanced Search
Filter by
Settings
Search Results
Item Embargo Adaptive AI-Based Enhancement of Critical External Sounds in Insulated Vehicle Cabins for Improved Safety(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Rathnayaka D.B; Wickramasuriya L.H.N.Y; Walpalage J.V; Rathnayake, SThe increasing acoustic insulation in modern and electric vehicles improves passenger comfort but unintentionally suppresses critical external sounds such as ambulance sirens, car horns, and train alarms, creating potential safety risks. While existing research has explored sound detection or localization in isolation, few systems integrate both capabilities in a unified framework for real-time vehicular deployment. This research proposes an adaptive AI-based system that detects, classifies, and selectively enhances these critical sounds in real time while providing directional awareness. Using a convolutional recurrent neural network (CRNN) trained on the UrbanSound8K dataset, the system processes incoming audio from external microphones, extracts Mel-frequency cepstral coefficients (MFCCs), and distinguishes safety-relevant cues from non-essential background noise. A dual-microphone setup enables the estimation of sound direction (left or right), providing additional spatial awareness to the driver. Detected signals are isolated through spectral filtering and relayed into the cabin with sub-30 ms latency, ensuring timely driver and passenger awareness without compromising comfort. Experimental results achieved 91.2% classification accuracy and 87.4% directional accuracy,confirming the system's feasibility for enhancing safety in insulated vehicle cabins and supporting future autonomous driving environments.Item Embargo Fidelity-Driven Physically Constrained Adaptive Code Distance for Surface Codes under Biased Circuit-Level Noise(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Hettiarachchi, A; Dissanayaka, KSurface codes are leading candidates for fault-tolerant quantum computing, but conventional implementations use fixed code distances that result in conservative resource allocation during varying noise conditions. We propose a physically-constrained, fidelity-driven adaptive surface code framework that dynamically adjusts code distance at scheduled intervals to maintain target logical fidelity while minimizing resource overhead. Our approach introduces a syndrome-based logical error rate estimation algorithm using rolling-window statistics, a scheduled adaptation policy operating within pre-allocated qubit pools, and detailed modeling of adaptation latency and reconfiguration-induced errors for superconducting qubit devices. Through large-scale Monte Carlo simulations under realistic biased circuit-level noise, we demonstrate that hardware-constrained adaptive codes achieve 43-62% success rate improvements over fixed-distance strategies, reaching 72-82% success rates compared to 5̃0% for fixed approaches. Oracle adaptive strategies with perfect error knowledge achieve 91-94% success rates, indicating substantial theoretical potential. Our results show that adaptive surface codes provide significant performance improvements even under realistic hardware constraints, with oracle strategies consuming only 37% of maximum resources while achieving 86% higher success rates for shorter circuits.Item Embargo Generative AI Based Chatbot for Customer IT Support Automation(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Lakshan S.N; Gamage U.R; Fernando W.S.N; Kuruppu K.A.V.U.; Kodagoda, N; Abeywardhana, LEnterprises are facing increasing challenges in delivering efficient, adaptive, and secure IT support, as traditional ticketing systems suffer from latency, rigidity, and limited scalability. Existing chatbot solutions built on commercial large language models (LLMs) often introduce high operational costs, token inefficiency, and data privacy concerns, while current retrieval and multi-agent systems lack personalization, multimodal understanding, and dynamic adaptability. This research presents a generative AI-based IT support assistant that integrates four core innovations: a multi-agent architecture enabling distributed and scalable task execution; a retrieval-augmented knowledge base enhanced with knowledge graphs and episodic memory for more accurate and personalized responses; conversational form automation using Rasa and transformer-based models to extract structured ticket information from natural queries; and an interactive 3D avatar interface with real-time speech-to-text, text-to-speech, and emotion-aware multimodal interaction to support more human-like communication. The system was evaluated using RAGAS metrics to measure accuracy, contextual recall, and response faithfulness. Results show improved domain-specific performance, reduced reliance on manual ticket handling, and lower computational overhead through caching and adaptive slot filling. This study proposes a privacy-preserving, scalable enterprise IT support framework, with future work targeting multilingual support and deeper ITSM integration.Item Embargo Real-Time Optimization and Maintenance of Wind Turbine Performance using Digital Twin Technology(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Jenojan P; Dhanushikan V; Herath H.M.T.S.; Dilmini N.A.C; Jayasinghearachchi V.; Perera JWind power plays a vital role in Sri Lanka's renewable energy transition, yet coastal wind farms face challenges such as lightning strikes, wind misalignment losses, turbine cut-in/out events, acoustic impacts, and costly maintenance. This study proposes a digital twin-based framework to enhance real-time optimization and predictive maintenance of wind turbine performance. The framework integrates four modules: weather risk forecasting, operational efficiency, noise impact analysis, and predictive maintenance. Using SCADA data from the Mannar Thambapavani Wind Farm, long-term meteorological and lightning records, NASA satellite observations, and the WEA-Acceptance dataset, advanced machine learning models were developed to forecast lightning (F1 = 0.81, AUC = 0.87), estimate power losses (R2 = 0.80, MAE = 3.4 kWh/h), optimize blade pitch, and produce short-term and medium-term energy forecasts. The digital twin simulation visualizes turbine dynamics, noise propagation, and maintenance scenarios. Results show improved prediction accuracy, 20-30% reduction in downtime, and a clear trade-off between energy efficiency (5-10° pitch) and noise (~56 dB > 25°). The framework strengthens situational awareness, operational reliability, and sustainability in monsoon-prone tropical environments.Item Embargo A Scalable Microkernel Inspired Architecture for Transparent and Adaptive Supply Chain Management(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Vithanage H.D; Dehipola H.M.S.N; Manditha K.D.R.; Weedagamaarachchi K.S.; Jayasinghearachchi, V; Perera, JSupply Chain Management (SCM) in industries such as coconut peat production increasingly demands systems that are not only efficient but also transparent and adaptable to rapid changes. Traditional monolithic architectures often fail to scale effectively and struggle to integrate emerging technologies such as IoT and blockchain, limiting their usefulness in dynamic environments. To address these limitations, this research proposes a microkernel-inspired architecture tailored for SCM. The architecture separates core system functions from domain-specific processes through the use of dynamically loadable plugins, allowing for modularity, fault isolation, and real-time adaptability. The impact of this study lies in its integration of multiple technologies to enhance transparency and operational flexibility. The system incorporates IoT sensors for real-time data acquisition, blockchain for immutable traceability, gRPC for high-performance communication, and K3S for lightweight container orchestration. A workflow customization tool allows non-technical users to define or modify supply chain processes without altering core logic. We employed the Architectural Trade-off Analysis Method (ATAM) and Cost-Benefit Analysis Method (CBAM) to evaluate architectural decisions and conducted runtime performance testing using simulated workloads. Results showed improved scalability, flexibility, and fault tolerance, with moderate latency introduced by inter-process communication and CGo overhead. Despite some limitations in performance variability, the architecture maintained high availability and was able to recover from plugin failures seamlessly. These findings suggest that the proposed model provides a viable foundation for next-generation SCM systems. The contributions include a modular, resilient framework that effectively integrates advanced technologies to support adaptable and transparent supply chain operations across various industries.Item Embargo Framework for Ethical Governance of Autonomous AI Systems: Balancing Innovation, Accountability, and Risk Management(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Dabare, C; Wijegunasighe, Y; Herath, H; Nethsara, R; Jayasinghe, PThe advent of rapid deployment of autonomous artificial intelligent systems is a twofold challenge: how to ensure this is happening in an ethical way, and at the same time, how to keep the innovative power. These are already outdated models of governance that render many of these models inadequate in accommodating the risk landscape as well as the applied complexity of autonomous artificial intelligent systems. This paper suggests a new conceptual model of governance that combines ethical standards, risk stratified regulatory feedback and allowing innovation via regulatory sandboxes. Based on a qualitative synthesis of worldwide AI policies, scholarly sources, and background information, the framework provides a hierarchical, modular paradigm that can be applied in different domains and jurisdictions. It focuses on transparency, accountability, fairness and incorporates flexibility of changing technology. The proposed model would play the role of a policymaker, industry practitioners, and researchers strategic road map in finding balanced and operationally feasible solutions in deploying ethical AI.Item Embargo HCLIP: Beyond CLIP for Cost-Effective Multimodal Retrieval in Education(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Weerasinghe, S; Gunatunga, O; Dewpura, W; Fernando, S; Kasthurirathna, D; Rathnayake, SMultimodal retrieval systems have gained significant attention due to their ability to process and cross-retrieve data containing images and text. However, the factors such as high cost of development, limitation on resources, and the proper addressing of the modality gap, the inherent representational differences between modalities pose a challenge to building effective and efficient retrieval models. In this work, we propose a low-resource, cost-efficient hybrid multimodal retrieval model that integrates Contrastive Language-Image Pre-training (CLIP) and All-MiniLM-L6-v2 to create a shared embedding space while storing raw images in an unstructured database. Our primary contributions include (1) the development of a hybrid model that outperforms CLIP-native retrieval, (2) a novel bidirectional neural network alignment technique that brings textual and visual modalities closer together, and (3) a comprehensive analysis of the modality gap's impact on downstream retrieval performance. Through proper evaluation using transparent techniques such as Mean Reciprocal Rank (MRR) and Cosine-Weighted MRR, our method demonstrates improved retrieval accuracy over baseline approaches. Experimental results exhibit that a lower modality gap does not always prove to be efficient on the downstream retrieval. Our findings pave the way for more efficient, adaptable, and cost-effective multimodal retrieval methodologies in low-resource environments, not limited to the education domain.Item Embargo EDU-SENSE: AI-Driven Adaptive E-Learning System for Primary Education(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Shanmugalingam, N; Maheshwaran, L; Rizvee, I; Roshni R.J; Samarakoon, U; Rajendran, KThis research addresses critical gaps in primary education by integrating AI-driven e-learning tools with traditional classroom teaching, ensuring technology supports rather than replaces teachers while improving student learning outcomes. In getting ready for high-pressure exams like Sri Lanka's Grade 5 Scholarship Examination, maintaining student engagement and emotional well-being remains a significant challenge. EDU-SENSE is an AI-driven web platform designed for primary education that integrates personalized content delivery, stress detection, and an emotion-aware chatbot. The system begins with a diagnostic pre-test to assess individual skill levels,and provides syllabus-aligned questions, and delivers targeted revision based on identified knowledge gaps. A s tress-aware module and motivational chatbot provide emotional support, while peer collaboration features promote teamwork and knowledge sharing. Using Machine Learning, Natural Language Processing, and fine-tuned Large Language Models, EDU-SENSE offers a holistic solution that enhances both academic performance and emotional well-being in young learners.Item Embargo Data-Driven Insights for Improved Diabetes Management(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Gunawardhana D.H.M.G; Kajeevan J; De Silva L.K.N; Wimansa P.P.H.S.D.; Anjana, J; Dassanayake, G.TEffective diabetes management requires understanding the complex interplay between multiple physiological and behavioral factors affecting glycemic control. This study presents a comprehensive analysis of diabetes management data from six distinct datasets containing continuous glucose monitoring (CGM) data, insulin administration records, carbohydrate intake, physical activity, sleep quality, and stress measurements. We employed time series analysis techniques such as STL decomposition, Dynamic Time Warping, LSTM neural networks, and XGBoost modeling to identify patterns in glucose variability and their relationships with modifiable factors. Our findings revealed significant time-of-day variations in glucose levels, with highest variability in late afternoon and evening (CV = 0.41 vs. 0.34, p < 0.01). Integrated prediction models achieved 83% sensitivity for hypoglycemia and 81% for hyperglycemia detection. The insulin-to-carbohydrate ratio varied by time of day, with morning ratios (1:8) differing from evening (1:12). Physical activity demonstrated intensity-dependent effects, with moderate-intensity exercise reducing glucose levels by 42 mg/dL for approximately 12 hours post-activity. Sleep quality below rating 3 (on a 5-point scale) was associated with a 37% increase in next-day glucose variability. These findings provide evidence-based insights for developing personalized diabetes management strategies that account for chronological variations in insulin sensitivity, meal timing, physical activity scheduling, and lifestyle interventions targeting sleep and stress management.Item Embargo An Approach to detect Advanced Persistent Threats using Machine Learning Techniques(Institute of Electrical and Electronics Engineers Inc., 2025-12-09) Bary A.A; Wijerupa W.D.O.D; Padukka P.V.G.G; Atapattu A.L.V.J.; Pandithage, D; Wijesooriya, AAdvanced Persistent Threats (APTs) pose significant risks to organizations due to their stealthy, prolonged nature and ability to evade traditional detection mechanisms. Traditional solutions often analyze separate data elements, such as network traffic or endpoint activity, limiting their effectiveness against sophisticated APT campaigns. This research proposes a holistic machine learning (ML)-driven approach to detect APTs by integrating three critical data dimensions: user behavior anomalies, endpoint activity monitoring, and network traffic analysis. The system further incorporates Tactics, Techniques, and Procedures (TTP) analysis using the MITRE ATT&CK framework to provide actionable intelligence. A real-time dashboard visualizes the threat detection results, TTP mappings, and mitigation strategies, enabling cybersecurity teams to respond proactively. The integration of multiple ML models enhances detection accuracy while bridging the gap between threat identification and contextual understanding. Experimental validation demonstrates the system's capability to detect APT indicators across diverse attack vectors and prioritize high-risk TTPs. This work contributes to advancing APT detection methodologies by offering a scalable, multi-dimensional solution tailored for modern cybersecurity operations.