The design and development of an Aerodynamic Ball Levitation Laboratory Plant, offering an engaging platform for studying control systems and real-time system identification. The system employs airflow to suspend a ping-pong ball within a vertical tube, counteracting gravitational forces to maintain the ball at a desired height. This dynamic setup presents a nonlinear, multi-variable control challenge, making it ideal for exploring control theory, system modeling, and stability analysis. A physical model of the Aerodynamic Ball Levitation system was developed alongside a mathematical model, and a classical PID controller was implemented to regulate the ball's position. PID parameters were systematically tuned through theoretical analysis and practical experimentation to achieve optimal control performance. MATLAB system identification tool was employed to model the system's dynamics accurately. The novelty of this work lies in its systematic and quantitative comparison of theoretical and real-time PID controller performance. Discrepancies in transient response metrics such as rise time, settling time, and overshoot were evaluated, identifying the impact of environmental disturbances and hardware limitations. The findings not only bridge the gap between theoretical design and practical application but also provide actionable insights into tuning strategies for real-time systems. These contributions advance the understanding of digital control systems in nonlinear applications and pave the way for further enhancements through adaptive control methods.

Publications

1. K. B. Wickramasingha, R. A. K. K. Perera and G. M. K. B. Karunasena, "Design and Analysis of Digital Control System Model for Aerodynamic Ball Levitation System," 2024 International Conference on Advances in Technology and Computing (ICATC), Kelaniya, Sri Lanka, 2024, pp. 1-7, doi: 10.1109/ICATC64549.2024.11025271.

Sarcopenia, defined by a gradual loss of muscle mass and function, affects many elderly people and is associated with reduced mobility, greater frailty, and a higher risk of falling. Using a vision attentive model and integrating the embedded Timed Up and Go Test (TUG-T), 3-meter Walk Test (3 mW−T), and fall risk analysis, this research proposes a unique method for assessing sarcopenia in elderly people. The attentive vision model uses computer vision techniques to examine TUG activities and gait speed in real-time, offering insightful information about the elderly’s the functional ability and muscular strength. Moreover, this approach provides a more comprehensive assessment of sarcopenia by integrating falling risk analysis. The proposed system achieved an overall accuracy of 86.6%, outperforming the individual components: TUG test (84.0%,p<0.05), gait speed (88.2%,p<0.05), and fallen risk assessment (93.0%,p<0.05). The results indicate that this novel strategy has enormous potential for aged healthcare, enabling targeted therapies and enhancing the overall quality of life for older people at risk of issues connected to sarcopenia.

Publications

1. H. M. K. K. M. B. Herath, A. G. B. P. Jayasekara, B. G. D. A. Madhusanka and G. M. K. B. Karunasena, "Non-Invasive Tools for Early Detection and Monitoring of Sarcopenia in Older Individuals," 2023 Moratuwa Engineering Research Conference (MERCon), Moratuwa, Sri Lanka, 2023, pp. 219-224, doi: 10.1109/MERCon60487.2023.10355475.

2. Herath, H.M.K.K.M.B., Jayasekara, A.G.B.P., Madhusanka, B.G.D.A., Karunasena, G.M.K.B. (2023). Attentive Vision-Based Model for Sarcopenia Screening by Automating Timed Up-and-Go (TUG) Test. In: Barsocchi, P., Parvathaneni, N.S., Garg, A., Bhoi, A.K., Palumbo, F. (eds) Enabling Person-Centric Healthcare Using Ambient Assistive Technology. Studies in Computational Intelligence, vol 1108. Springer, Cham. https://doi.org/10.1007/978-3-031-38281-9_4 

Development of the Solar Remote Laboratory for the Department of Mechanical Engineering, OUSL initiated under the EUSL energy program funded by European Union in 2020. Solar remote laboratory intended to carry out experiments based on control and real environmental conditions to cater to different learning outcomes. The solar remote laboratory is a setup that could perform experiments and gather data for research purposes remotely controlled by the experiment setup.

Development Group

1. Mr. HDNS Priyankara

2. Mr. GMKB Karunasena 

Nowadays, individuals and communities are benefiting from early recognition of stress. The American Psychological Association states that 75% of individuals are suffering moderate to severe levels of stress. Traditional stress detection methods rely on physiological signals, which are contact-based and require sensors to be in close proximity to humans. As a result, developing a reliable method of detecting stress that does not rely on physiological signaling is still a challenge. Several researchers in the field have used facial expressions and physiological signals to indicate stress levels. Stress might manifest itself in the form of facial skin conditions in the severe stage. The American Institute of Stress states that stress causes damage to the facial skin in different ways. There is, however, a gap in the field for describing stress using a combination of valence, arousal, and physical changes caused to the face. Contribution: This research is evaluating the relationship between mind (stress) and skin (facial skin) using factors such as valence, arousal, and facial skin conditions using the affective computing approach. Results: We investigated the mind-skin connection between 37 emotional test subjects (mean age: 31 ± 4 yrs.), and observed that V-A evaluation was 80.83% accurately showed promising results. The stress test group had a direct association with their facial skin, according to our findings. Moreover, 85.5% of the stress test group were suffered from high to server-level of skin conditions. 

Publications

1. Herath, H.M.K.K.M.B., Karunasena, G.M.K.B., Mittal, M. (2022). Monitoring the Impact of Stress on Facial Skin Using Affective Computing. In: Mittal, M., Goyal, L.M. (eds) Predictive Analytics of Psychological Disorders in Healthcare. Lecture Notes on Data Engineering and Communications Technologies, vol 128. Springer, Singapore. https://doi.org/10.1007/978-981-19-1724-0_4 

This research paper concerns the machine learning approach for tea bud leaf identification. The tea bud identification is most important for the process of automated tea leaves grading machines. In the present situation, there are no methods to identify the tea bud leaf separately from the main tea leaf. Unfortunately developing of mechanism for identification process is impossible because the plucked tea leaves not in the same condition. Therefore, the identification is needs intelligent practice to detect the tea bud leaf. In this research, the machine learning object detection technique is developed and successfully used for identify tea bud leaf and the capability of this proposed technique is validated through experimental results obtained by performing experiments by using MATLAB software. According to results, the proposed methodology provides 55% of overall accuracy for identification.

1. Karunasena, G. M. K. B., and Priyankara, H. D. N. S. (2020). Tea bud leaf identification by using machine learning and image processing techniques. Int. J. Sci. Eng. Res. 11, 624–628. doi: 10.14299/IJSER.2020.08.02

This paper concerns the control behavior of a magnetic levitation system using a fuzzy logic controller (FLC). The magnetic levitation system is a feedback control system already acknowledged and accepted in advance control systems. This kind of system needs a suitable intelligent controller for positioning a metal sphere in airspace with the help of an electromagnetic force. The magnetic force produced by an electromagnet and magnetic force controlled by changing the supply current. The position of the metal sphere controlled by changing the electromagnetic force. FLC proposed in this paper is to control the magnetic levitation system, developed by using Takagi-Sugeno fuzzy controller. The final simulation results with different initial conditions concerning the ball's desire and actual responses are provided to validate the theory. The capability of this proposed technique is to validate through experimental results obtained by performing experiments by using a real-time magnetic levitation system. Based on the experimental results and response analysis, it shows the fuzzy controller can stabilize the system and efficiently follow the desired trajectory the same as the PID controller response.

Publications

1. G.M.K.B. Karunasena, H.D.N.S. Priyankara,B.G.D.A. Madhusanka (2019). Design andDevelopment of Fuzzy Logic Controller for MagneticLevitation System. Conference: InternationalConference on Artificial Intelligence, SLAAI-ICAI-2019 (PDF) Artificial Neural Network vs PID Controller for Magnetic Levitation System. 

Recently, there has been a much public concern over fake photos and films that contain facial information created by digital modifications, especially with deepfakes techniques. The widely used term “Deepfakes” describes a deep learning technique that swaps faces to produce fake identities. The distinction between actual and deepfake videos has become crucial due to their detrimental effects on the world. An intriguing vulnerability within live deepfakes has largely gone unnoticed by the security research community thus far. This research may be attributed to real-time deepfakes during video calls and conferences, which have recently become a notable cause for concern. Developing and evaluating existing deepfake detection methods requires large-scale datasets. Still, the deepfake datasets available today have inadequate image quality and do not appear like the deepfake videos that have been making the rounds on the Internet. This research presents a novel way for recognizing synthetic media produced using deepfake: an active and passive facial features based end-to-end deep neural network classifier model (AFPF-DNNC model). The facial landmark detector recognizes and extracts the area from each individual video frame that contains the human face, then extracted facial regions used to extract active features in the faces. Simultaneously, a convolutional neural network (CNN) extracts passive features from these faces. The retrieved features are inputs into the end-to-end deep neural network classifier (DNNC), which is the last recognition step in the detection system. The proposed algorithm achieves 90.32% of an area under the receiver operating characteristic curve (AUC), 90.6 % accuracy, 90 % specificity, 91.3 % recall, 87.5 % precision, and 89.3 % F-score on the Celeb-DF (v2) a new deepfake forensics dataset. Finally, the experimental investigation validates the suggested method’s effectiveness over the most advanced techniques.

Publications

1. Karunasena, G.M.K.B., Herath, H.M.K.K.M.B., Rathnayake, R.M.P.M.D., Priyankara, H.D.N.S., Madhusanka, B.G.D.A. (2024). Detecting Deepfakes Through the Classification of Facial Active and Passive Features Using Machine Learning. In: Hewage, C., Yasakethu, L., Jayakody, D.N.K. (eds) Data Protection. Springer, Cham. https://doi.org/10.1007/978-3-031-76473-8_10