Poster Award

Takako Nonaka (Shonan Institute of Technology, Japan)

This study presents the design and development of the SQC-03, an electric-assist four-wheel bicycle intended for secondary transportation in the Shonan area. Addressing the limitations of its predecessor (SQC-02), the SQC-03 is designed to be lighter, more maneuverable, and suitable for both men and women. A posture verification experiment optimized the relationship between the handlebars, pedals, and seat for a comfortable riding posture. Key improvements include a reduced frame weight, spoke-type wheels, a shortened wheelbase for better turning, and an adjustable saddle position. The SQC-03 meets standard bicycle requirements while significantly improving maneuverability and accessibility. Future studies will evaluate long-term driving performance and ride comfort through extensive testing.

Wai Yie Leong (INTI International University, Malaysia)

The integration of Artificial Intelligence (AI) in classroom monitoring is transforming traditional education by enabling real-time analysis, personalized feedback, and automated oversight of learning environments. AI-powered classroom monitoring systems leverage computer vision, natural language processing (NLP), machine learning, and Internet of Things (IoT) sensors to track student engagement, behavior patterns, and academic progress. These systems help educators enhance instructional methods, optimize classroom management, and identify students who need additional support. This paper explores the evolution, methodologies, and applications of AI in classroom monitoring, examining key technologies such as facial recognition for attention tracking, sentiment analysis for emotional assessment, automated speech recognition (ASR) for participation analysis, and AI-driven adaptive learning. Through comparative studies, the research highlights the advantages of AI-powered monitoring over traditional classroom management techniques, showcasing improvements in student engagement, teacher efficiency, and learning outcomes.

Rick Lin (Morrison Academy, Taiwan); Iuon-Chang Lin (Chung Hsing University, Taiwan)

As online services expand, secure user authentication is essential. Traditional password-based authentication evolved into two-factor authentication (2FA) using OTPs, but security concerns led to the adoption of Public Key Infrastructure (PKI). However, issues such as device management and certificate renewal limit PKI's practicality for widespread use. OpenID authentication simplifies this by enabling users to verify their identity through a trusted provider, issuing ID_Token for authentication and Access_Token for service access. This paper explores Mobile Network Operator (MNO) authentication systems, particularly the Mobile Connect framework. The system employs a two-factor approach: (1) identity verification via government-issued documents during MSISDN registration and (2) authentication reinforcement through IMS registration (EPS-AKA), establishing a unique MSISDN-IP relationship. User data is encrypted, validated via telecom APIs, and cross-verified with telecom databases and SIM card identity records. Analyzing existing MNO authentication mechanisms, this research identifies security vulnerabilities in web interfaces and introduces an IP_Token mechanism to enhance security. This approach improves token management, minimizes user data exposure in API responses, and ensures compliance with legal and auditing requirements by integrating authentication records into a secure database.

Ming-Lin Chuang and Ming-Tien Wu (National Penghu University of Science and Technology, Taiwan)

This work presents a 2x2 polarization-diversity patch array antenna with a compact feeding circuit. Using the corresponding inputs, the designed array antenna can generate six different polarization states. The performance of the proposed array antenna is the same as the conventional 2x2 array antenna with a complicated feeding circuit. Therefore, the proposed compact polarization-diversity antenna could be used in future wireless communications.

Hideya So (Shonan Institute of Technology, Japan); Jin Nakazato (The University of Tokyo, Japan)

UAV delivery services can help address the shortage of truck drivers and enable blind flight in urban areas, making urban delivery more feasible.
However, since UAVs operate in the sky, their radio waves reach farther than ground-based communications, increasing radio interference.
The authors evaluated this interference using a 3D city model from PLATEAU and ray-traced simulations.
This paper analyzes radio interference in Tokyo, Japan, based on actual user distribution.
It reveals that the percentage of affected users depends on the number of users near UAVs.

Wai Yie Leong (INTI International University, Malaysia)

Autonomous game balancing through AI-driven dynamic difficulty adjustment (DDA) is a cutting-edge approach that leverages artificial intelligence to optimize gameplay experiences dynamically. This paper explores various AI-driven DDA techniques, including supervised learning, reinforcement learning, and hybrid models, to ensure player engagement and fairness. By analyzing fairness metrics such as the Gini coefficient, win rate distribution, and player retention rates, we evaluate the impact of AI-driven balancing on gaming experiences. A case study involving a multiplayer shooter game demonstrates the effectiveness of AI-based difficulty tuning, showcasing improvements in win rate consistency, player engagement, and skill gap reduction. The study also highlights key challenges, including computational overhead, AI bias, and transparency issues. Future research directions focus on explainable AI, ethical frameworks, and fairness-aware algorithms to enhance AI-driven game balancing further.

Chang-Long Jiang, Yi-Kuang Tsai, Shih-Hsiung Lee, Cheng-Che Hsueh and Ko-We Huang (National Kaohsiung University of Science and Technology, Taiwan)

In recent years, heuristic algorithms have been increasingly utilized to support decision-making and parameter optimization. For instance, in investment portfolio optimization, they are employed to identify the optimal stock combinations that balance risk and return. This paper introduces a stock price prediction system, AVLA (AOA with Associated Network), which integrates the Arithmetic Optimization Algorithm (AOA), Variational Mode Decomposition (VMD), Long Short-Term Memory (LSTM), and an associated network architecture. Experimental results demonstrate that AVLA outperforms baseline models in stock price prediction, highlighting its effectiveness in financial forecasting

Chien-Lin Chen and Cheng-Yuan Chang (National United University, Taiwan)

In this paper, the design and implementation of apple quality automatic classification system based on deep learning (DL) technique is proposed. According to the quality classification experience of fruit farmers, this paper trains a DL model to classify the apples into three levels (i.e., grade A, grade B, and grade C), which represent highest-quality apples, fair-average quality apples, and lowest-quality apples, respectively. The training model is used to construct a complete apple quality classification system which mainly consists of a conveyor, a robotic arm, a webcam, microcontroller unit (MCU) ESP32 and so on. The experiment results show that the proposed system can accurately classify the apples to the correct corresponding packaging areas indeed, thus not only increasing the reliability of classification, but also improving the work efficiency of the fruit farmers, especially in the aging of the agricultural labor force.