This paper proposes a flux saturation model of synchronous reluctance machines (SynRMs) using an artificial neural network. The proposed ANN model is structured with current as the input and flux as the output, and supervised learning is conducted using the current and flux data. Considering test conditions of motor drive systems, flux data were acquired through two methods based on the voltage equation. For the first one, flux data were obtained in a stationary state using hysteresis current control. In the other, flux data were acquired in a rotating state. Since the acquired data are influenced by sensor noise and inverter non-linearity, data preprocessing was performed through cleansing and normalization. Furthermore, the model was designed to be light-weighted for applications in motor drive systems. To verify the feasibility of two ANN models which were trained with different datasets, the outputs of the ANN models were compared with the current-flux data obtained in the rotating state. Moreover, the proposed ANN models were verified using model-based sensorless drives in a 1.5 kW SynRM.

The rapid growth of electric vehicles (EVs) in recent years has been propelled by their eco-friendly benefits and expanding role within sustainable energy systems. In smart grids, EVs act as flexible electrical loads and a distributed energy storage unit. Using vehicle-to-grid (V2G) technology, they exchange power bidirectionally to enhance microgrid flexibility and increase renewable-energy utilization. Leveraging time-varying electricity prices and V2G capabilities enables microgrid operators to cut operational costs by adjusting EV charge–discharge schedules. However, designing these optimal schedules is challenging, as EV arrival and departure times, as well as electricity price trajectories, are highly unpredictable. To address this issue, the scheduling problem was modeled within a Markov Decision Process (MDP) framework, followed by the development of a deep reinforcement learning (DRL)-based solution. Specifically, an algorithm based on a Twin-delayed deterministic policy gradient (TD3) is implemented to autonomously derive optimal continuous charging and discharging decisions through real-time interactions with the environment. Extensive simulations demonstrate that the proposed DRL-based framework effectively reduces microgrid operational costs while ensuring EV owner satisfaction through State of Charge (SoC) requirements, optimal use of photovoltaic (PV) energy, and coordinated real-time V2G scheduling across all vehicles.

This paper addresses the rendezvous problem in asymmetric cognitive radio networks (CRNs) by proposing a novel reinforcement learning (RL)-based channel-hopping algorithm. Traditional methods like the jump-stay (JS) algorithm, while effective, often struggle with high time-to-rendezvous (TTR) in asymmetric scenarios where secondary users (SUs) have varying channel availability. Our proposed RL-based algorithm leverages the actor-critic policy gradient method to learn optimal channel selection strategies by dynamically adapting to the environment and minimizing TTR. Extensive simulations demonstrate that the RL-based algorithm significantly reduces the expected TTR (ETTR) compared to the JS algorithm, particularly in asymmetric scenarios where M-sequence-based approaches are less effective. This suggests that RL-based approaches not only offer robustness in asymmetric environments but also provide a promising alternative in more predictable settings.

This study proposes artificial intelligence (AI) technology for the automatic classification of Korean scientific and technical papers, aiming to achieve high accuracy even with a small amount of labeled data. Unlike existing BERT-based Korean document classification models that perform supervised learning based on a large amount of accurately labeled data, this study proposes a structure that utilize large language models (LLMs) and retrieval-augmented generation (RAG) technology. The proposed method experimentally demonstrates that it can achieve higher accuracy than existing technologies across all cases using various amounts of labeled data. Furthermore, a qualitative comparison between manually-generated labels, and recognized as correct answers and those produced by LLM responses confirmed that the LLM responses were more accurate. The findings of this study, while limited to Korean scientific documents, provide evidence that a system utilizing LLM and RAG for document classification can easily be extended to other domains with diverse document datasets, owing to its effectiveness even with limited labels.

The performance of a frequency-hopping spread-spectrum system is mainly dependent on the mathematical properties of its hopping sequences, which are designed to minimize interference between different users. The one-coincidence sequence frequency-hopping sequence (OC-FHS) set is one of the primary types, because it achieves the lowest possible values regarding Hamming auto- and cross-correlation. In this work, we propose an OC-FHS set of a prime length p and alphabet size $pq$ for two primes p and q using a block structure modulo $pq$. In particular, when $p=q$, our construction provides a significantly larger set size compared with a previously known OC-FHS set with the same length and the same alphabet size. Moreover, the set size is optimal with respect to the bound established by Cao, Ge, and Miao. This extended set size can be applied to FHMA systems that need to accommodate a large number of users.

Industrial control systems must support both periodic and aperiodic traffic, each with distinct timing and reliability constraints. In particular, aperiodic traffic is classified into three priority levels based on time urgency. While periodic tasks follow fixed schedules, aperiodic tasks are event-driven and require either deadline-aware access or low latency. To address these heterogeneous quality of service (QoS) requirements in duty-cycled, bandwidth-constrained long range (LoRa) networks, we propose QoS-LoRa, a deterministic MAC protocol designed for mixed-criticality traffic. QoS-LoRa integrates three key mechanisms: logical frame partitioning with distributed contention slots, deadline-aware contention window scheduling, and delay-slot-based prioritized access. Together, these components ensure low collision probability, bounded delay, and priority-respecting access across different class tasks. Simulation results show that QoS-LoRa achieves over 95% packet delivery ratio (PDR) for deadline-constrained traffic and over 92% for best-effort traffic, even under 60% aperiodic load. QoS-LoRa significantly outperforms RT-LoRa-LFP and ALOHA-based baselines in PDR, while maintaining end-to-end delay comparable to RT-LoRa-LFP, validating its effectiveness in industrial environments.