International Journal on Robotics, Automation and Sciences

An Edge Convolution Neural Network Model for Plant Health Classification Using Camera

2023-06-21T10:13:12+08:00

As per the Food and Agricultural Organization (FAO), plant diseases infect approximately 1.3 billion tonnes of crops. Historically, farmers relied on visual inspection for disease detection and classification. In this study, a Convolutional Neural Network (CNN) with five convolutional layers was used to accurately recognize plant diseases. A deployable CNN model was developed for classifying plant diseases, integrated into a web application with a camera, forming a vision system integrated with CNN model. The CNN model was trained using a public dataset comprising 19,384 images of potatoes, peppers, and tomatoes, collected under controlled conditions. These plants were chosen due to their common occurrence in Malaysia. The evaluation metrics F1 score were used to assess the model’s performance. The accuracy and F1-score of the trained model were 97.2% and 97%, respectively.

Manuscript received: 26 Nov 2024 | Revised: 3 Jan 2025 | Accepted: 11 Jan 2025 | Published:: 31 Mar 2025

Development of an IoT-based Wireless Controlled Power Adapter

2023-08-01T10:07:06+08:00

Energy waste issues in electrical appliances due to inefficient usage monitoring are commonly faced by almost every person. This project aims to develop a wireless controlled power adapter operates autonomously based on occupancy of a specific space. In this way, convenience is brought to the user, and energy waste could be prevented. This system provides two modes of operation: manual and automatic. Using the mobile phone user interface, the user can manually and wirelessly control the power adapter. When there are no occupants in the specific space, the system will automatically shut off the power adapter. In contrast, if a person is detected in a specific space, the power adapter will be automatically switched on. WIFI protocol is used for the entire communication system. Experimental demonstration has been conducted to show the functionality. This proposed system offers the user to control in a long-distance range as long as the system is connected to a random WIFI network.

Manuscript received: 27 Dec 2024 | Revised: 21 Jan 2025 | Accepted: 13 Feb 2025 | Published: 31 Mar 2025

From Signatures to AI: A Comprehensive Review of DDoS Detection Strategies in IoT & SDN

2024-02-07T12:00:15+08:00

In the ever-evolving landscape of the Internet of Things (IoT) and Software-Defined Networks (SDN), the rapid growth of interconnected devices has enhanced ease and efficiency. However, this evolution has also paved the way for the ominous cyber-attack: Distributed Denial of Service (DDoS). These attacks, which make systems unavailable for legitimate users, threaten the data integrity, confidentiality, and availability in IoT and SDN infrastructure. This paper delves into the critical issue of DDoS attacks within the IoT and SDN environments, offering a comprehensive exploration of detection mechanisms by categorizing them into traditional (signature-based) and anomaly-based approaches i.e., Machine Learning (ML), Deep Learning (DL), and statistical techniques. Our key findings reveal that while signature-based methods effectively identify known attack patterns, they fall short against novel threats. In contrast, AI-based approaches, particularly ML and DL, demonstrate superior performance in detecting previously unseen attacks. However, their efficiency is highly dependent on the quality of training data and model robustness. Our comparative analysis indicates that ML and DL methods achieve higher detection rates and lower false positives in experimental settings, underscoring the importance of high-quality datasets and resilient models. By highlighting the strengths and limitations of both approaches, this study provides valuable insights for researchers and cybersecurity experts. The need for an effective and diversified DDoS detection mechanism in the developing IoT and SDN domains is evident. While conventional methods remain relevant, AI-based strategies offer a dynamic avenue for enhancing security.

Manuscript received: 24 Oct 2024 | Revised: 14 Dec 2024 | Accepted: 30 Dec 2024 | Published: 31 Mar 2025

Vendor Evaluation and Selection for Forwarding Activities Using Stepwise Weight Assessment Analysis-Combined Compromise Solution (SWARA-CoCoSo) Method

2024-04-07T11:39:45+08:00

PT Perusahaan Listrik Negara Suku Cadang has faced delays in receiving commodities requested by the Perusahaan Listrik Negara Group, resulting in materials arriving later than expected. These materials were supplied by PT Wartsila, a partner of PT Perusahaan Listrik Negara Suku Cadang, responsible for fulfilling the orders placed by the Perusahaan Listrik Negara Group. To uphold its reliability as a supply chain company, PT Perusahaan Listrik Negara Suku Cadang must ensure timely delivery of requested goods. One way to minimize delays is through vendor evaluation. The SWARA method, which assesses ten factors identified by four logistics division experts, is employed to select the best forwarding vendor. The CoCoSo method, along with PT Perusahaan Listrik Negara Suku Cadang's logistics performance evaluation, was used to determine the top vendor. Based on the CoCoSo results, PT Kurnia Purnama Jaya ranked first with a score of 4.06, followed by Mats International Indonesia with a score of 2.6, Perigi Raja Terpadu in third with 1.5, and Pos Logistik Indonesia in fourth with 1.4. According to the CoCoSo method’s criteria weightings and vendor evaluation, PT Kurnia Purnama Jaya was selected as the most suitable vendor for PT Perusahaan Listrik Negara Suku Cadang.

Manuscript received: 19 Sep 2024 | Revised: 12 Dec 2024 | Accepted: 29 Dec 2024 | Published: 31 Mar 2025

Exploring Activities of Daily Living Among the Elderly through Machine Learning Techniques

2024-08-22T12:18:36+08:00

Activities of daily living (ADLs) is a term that is used to describe the activities performed in everyday life that involves the motion of the human body such as eating, walking, and sitting. ADLs can be used to determine the state of elderly people as a decline in ADL performance will generally mean a decline in the human body. It can act as an early indicator if an elderly person is experiencing underlying illness or health issue. This project aims to detect five different ADLs which are eating, cooking, sweeping, walking, and sitting and standing. A dataset was collected from twenty individuals performing each ADL at two different angles, a front view and a side view. A computer vision-based human pose estimation technique is used to extract the human body keypoints. These keypoint values are then processed and fit into multiple deep learning models for analysis. In this study, five different deep learning models namely LSTM, Bi-LSTM, CNN, RNN and Transformer models have been evaluated. The performance of each model is analysed and discussed. It was determined that the CNN model performed the best achieving a categorical accuracy of 82.86%.

Manuscript received: 14 Sep 2024 | Revised:4 Dec 2024 | Accepted: 11 Dec 2024 | Published: 31 Mar 2025

Developing a Telepresence Robot for Autism Diagnosis

2024-08-14T17:28:14+08:00

The global COVID-19 pandemic posed significant challenges to the healthcare industry in maintaining continuous operations while adhering to strict physical distancing protocols. Critical functions such as delivering meals to patients, supplying medical instruments, monitoring vital signs, assisting those with impaired mobility, and ensuring accurate disease diagnoses became increasingly difficult. As the world adapts to a post-pandemic reality, robots are expected to play a more prominent role by becoming more self-reliant, adaptable, and collaborative. In response to these evolving needs, the Centre for Unmanned Technologies (CUTe) at International Islamic University Malaysia (IIUM), in collaboration with Prostrain Technologies, developed the innovative medical robot called "Medibot". Medibot, a telepresence robot, presents a promising tool for observing children's true behaviours and interactions—essential for diagnosing Autism Spectrum Disorder (ASD). Equipped with a high-resolution camera, Medibot facilitates seamless video conferencing between children and experts, enabling detailed behavioural analysis during diagnostic sessions. The presence of parents beside the child enhances comfort, while the robot's non-intrusive character encourages natural responses and interactions. Compared to traditional human-led assessments, Medibot's presence is less intimidating, potentially leading to more accurate diagnoses. Medibot’s development is underpinned by a robust ROS-based software architecture, enabling autonomous navigation in complex hospital environments while avoiding static and dynamic obstacles with high operational consistency. Extensive testing has validated its mapping and navigation capabilities, ensuring smooth and predictable movements without human intervention, making the diagnostic process less intrusive and seamless. The incorporation of telepresence technology, primarily through a teleconferencing camera for live image streaming, represents a significant advancement in remote healthcare. With applications ranging from ASD diagnosis to broader medical monitoring, Medibot exemplifies the transformative potential of telepresence robotics in expanding access to specialized care and improving patient outcomes.

Manuscript received: 14 Sep 2024 | Revised: 18 Dec 2024 | Accepted: 3 Jan 2025 | Published: 31 Mar 2025

Review on Advancements in Artificial Intelligence and its Applications in Sports

2024-11-19T11:23:34+08:00

The sport industry is being transformed by Artificial Intelligence (AI) in many ways. This paper seeks to discuss how AI has improved sports science, particularly in boosting the athletes’ performance and avoiding injuries, through various machine learning models like Extreme Gradient Boosting, Support Vector Machines, and Random Forest Regression. These AI tools are more effective than the traditional methods, as they predict the athletes’ performance results more accurately and managing their injuries more proactively. This paper also discusses the challenges of using AI in the sport industry, particularly in terms of data privacy and the reliability of the models. With the aid of AI, it is of no doubt that sport science will have a promising future.

Manuscript received: 24 Oct 2024 | Revised: 10 Dec 2024 | Accepted: 17 Dec 2024 | Published: 31 Mar 2025

Improve Exercise Movement: Detecting Mistakes on Yoga with Mediapipe and MLP

2024-11-05T13:10:08+08:00

Yoga is known as a comprehensive practice for maintaining physical and mental health. However, improper execution of yoga postures can cause injury, hinder progress, and potentially damage health. To overcome this problem, this research utilizes Mediapipe as a data preprocessing tool to identify yoga poses, which are then classified using the Multi-Layer Perceptron (MLP) algorithm. In the process, data normalization is carried out to increase prediction accuracy. This research uses a dataset consisting of six classes of yoga poses, namely tree, downdog, goddess, warrior, and plank. Experimental results show that the model achieved 98% accuracy during training, but accuracy during testing decreased to 95%. This shows an indication of overfitting, where the model adapts too much to the training data and is less able to generalize to the test data. This study makes an important contribution to the development of a safer and more accurate yoga pose classification system, which can be applied to practice yoga properly and prevent injuries.

Manuscript received: 15 Oct 2024 | Revised: 30 Jan 2025 | Accepted: 11 Feb 2025 | Published: 31 Mar 2025

Enhancing LLM Efficiency: A Literature Review of Emerging Prompt Optimization Strategies

2024-11-11T16:50:35+08:00

This study focuses on enhancing the performance of Large Language Models (LLMs) through innovative prompt engineering techniques aimed at optimizing outputs without the high computational costs of model fine-tuning or retraining. The primary objective is to investigate efficient alternatives, such as black-box prompt optimization and ontology-based prompt refinement, which improve LLM performance by refining prompts externally while maintaining the model's internal parameters. The study explores various prompt optimization techniques, including instruction-based, role-based, question-answering, and contextual prompting, alongside advanced methods like CoT and ToT prompting. Methodologically, the research involves a comprehensive literature review, benchmarking prompt optimization techniques against existing models using standard datasets such as Big-Bench Hard and GSM8K. The study evaluates the performance of approaches like APE, PromptAgent, self-consistency prompting, and many more. The results demonstrate that these techniques significantly enhance LLM performance, particularly in tasks requiring complex reasoning, multi-step problem-solving, and domain-specific knowledge integration. The findings suggest that prompt engineering is crucial for improving LLM efficiency without excessive resource demands. However, challenges remain in ensuring prompt scalability, transferability, and generalization across different models and tasks. The study highlights the need for further research on integrating ontologies and automated prompt generation to refine LLM precision and adaptability, particularly in low-resource settings. These advancements will be vital for maximizing the utility of LLMs in increasingly complex and diverse applications.

Manuscript received: 3 Oct 2024 | Revised: 13 Dec 2024 | Accepted: 25 Dec 2024 | Published: 31 Mar 2025

A Review of Camouflage Object Detection Techniques

2025-01-03T15:31:19+08:00

Camouflage Object Detection (COD) is a constantly evolving field that deals with the difficulties of locating items hidden in intricate settings. This review examines the progression of COD techniques, from classical human methods to physical component-based methods such as infrared, LIDAR, multispectral and hyperspectral detection. Key applications of COD span from military reconnaissance to wildlife monitoring, medical imaging, and disaster response, where the ability to detect concealed objects has transformative implications. Future research should prioritize integrating diverse data sources, refining machine learning algorithms, and overcoming deployment constraints to advance the field further.

Manuscript received: 30 Dec 2024 | Revised: 30 Jan 2025 | Accepted: 17 Feb 2025 | Published: 31 Mar 2025

The Role of Electroencephalography in Advancing Sleep Research

2025-01-23T12:52:41+08:00

Electroencephalography (EEG) is fundamental in sleep research, providing critical insights into cerebral activity and significantly contributing to the diagnosis of sleep disorders. This study examines recent progress in EEG-based sleep research, emphasizing cutting-edge methods for sleep staging and disease identification. The amalgamation of machine learning and deep learning methodologies, encompassing hybrid models such as CNN-LSTM, has markedly improved the precision of sleep stage categorization and automated analysis. Enhancements in signal quality and dependability, especially by improvements in artifact removal methods like wavelet-enhanced independent component analysis (ICA), have further advanced these developments. The implementation of multimodal strategies, wearable EEG technology, and AI-enhanced systems has broadened the sphere of sleep monitoring beyond clinical environments, rendering it more accessible and individualized. This article examines the use of EEG in detecting sleep disorders, including insomnia, obstructive sleep apnea, and narcolepsy, by identifying biomarkers and abnormalities in sleep architecture. Emerging research underlines the promise of clinical EEG, marking it as a transformational tool for both study and therapy. Nonetheless, obstacles persist in domains such as noise reduction, biomarker standardization, and scalability. Future directions include merging EEG with imaging modalities like fMRI, developing wearable technology, and employing advanced AI for individualized sleep health management. In particular, EEG is highlighted as a transformational and promising tool for promoting sleep medicine through novel, accessible, and effective solutions.

Manuscript received: 5 Jan 2025 | Revised: 10 Feb 2025 | Accepted: 18 Feb 2025 | Published: 31 Mar 2025

Real-Time Emotion Detection Using Artificial Intelligence: A Review

2025-02-07T13:38:40+08:00

The integration of artificial intelligence (AI) in emotion recognition has significantly transformed human-computer interaction and revolutionized fields such as medicine, education, and entertainment. This paper reviews 30 papers on the detection of emotional signs through various biometric inputs, including electroencephalography (EEG), electrocardiography (ECG), facial expressions, and speech patterns. Despite advancements in AI-driven emotion recognition systems, challenges persist, particularly in data variability, computational inefficiency, and ethical dilemmas associated with privacy, security, and algorithmic bias. Recent innovations in feature extraction techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), have enhanced the precision of emotional state recognition across multiple input channels. The transition to edge computing has further enabled real-time processing with low latency, facilitating integration into wearable devices and IoT ecosystems. Multimodal systems, which leverage data sources such as physiological signals, facial expressions, and speech, show great promise but face challenges related to inclusivity and system fragility. To address these issues, the study recommends for robust training datasets, ethical guidelines, and hardware optimizations. Incorporating contextual information and accounting for individual differences can improve recognition accuracy and user trust. However, ethical concerns remain critical, emphasizing the need for strict standards of privacy, security, and equitable access to ensure AI emotion recognition systems are trustworthy and inclusive. Overall, this paper highlights the potential of AI-driven emotion recognition systems while underscoring the importance of continuous research to address technical and ethical challenges, paving the way for broader applications in pattern recognition, cognitive studies, and specialized tools.

Manuscript received:4 Jan 2025 | Revised: 13 Feb 2025 | Accepted: 27 Feb 2025 | Published: 31 Mar 2025

AI-Assisted Analysis for Breast Cancer Imaging and Diagnostics

2025-02-07T13:58:14+08:00

Breast cancer cases have increased by 0.5% each year. X-ray, CT-Scan, and magnetic resonance imaging have been used to detect cancer without harming the patient. However, these methods usually used manual screening to process medical images, which leads to longer processing time and increases the burden on the expert. With the help of deep learning, automation-driven breast cancer detection, segmentation, and explanation can be performed in the process, which can greatly reduce the processing time and the burden on experts. This paper proposed a deep learning model, S-YOLOv11 by combining YOLOv11 with SimAM attention mechanism and a GUI with integration of a large language model. The model is trained with 624 images and tested with 156 images. Several YOLO architectures were compared, including YOLOv8, YOLOv9, YOLOv10, and YOLOv11. The proposed model has outperformed the other models. In the detection task, 0.806 precision, 0.635 recall, and 0.724 mAP were achieved. In the segmentation task, 0.833 precision, 0.65 recall, and 0.739 mAP were achieved. In addition, the study also improved the functionality of the GUI by accessing the ChatGPT API. It is possible to generate medical analysis for breast cancer tumors, with the use of GUI for visualization. However, current research is still in the development stage and it needs to be put into clinical trials before it can be used.

Manuscript received: 3 Jan 2025 | Revised: 20 Feb 2025 | Accepted: 27 Feb 2025 | Published: 31 Mar 2025