Introduction: To address the inefficiencies and high costs of manual inspection in high voltage converter stations, this paper proposes an improved A* path planning algorithm. Methods: To overcome A*'s limitations (single-objective heuristic, inefficient search, vague node selection), this paper introduces three improvements: (i) a multi-objective cost function balancing energy, time, dynamics, and coverage with hierarchical constraint handling; (ii) adaptive 4/8- direction search based on obstacle distribution to reduce redundancy; (iii) a two-level open-list tiebreaker (min h(n), then min g(n)). Experiments use a 50×50 grid map. Results: Compared with traditional A*, the proposed algorithm reduces the number of expanded nodes by 32.2%, planning time by 26.9%, and the comprehensive cost (energy + time + dynamic) by 16.1%, while achieving 100% sampling point coverage. Field applications further validate its effectiveness in the DC yard and filter yard of a converter station. Discussion: The improved algorithm demonstrates significant advantages in reducing computational load (32.2% fewer nodes, 26.9% faster planning) and path cost (16.1% improvement) while maintaining complete coverage. The adaptive neighborhood search and multi-objective cost function contribute to these gains by reducing redundant turns and balancing multiple operational constraints. Field validation confirms that the improvements translate to real-world robotic inspection tasks in complex environments. Conclusion: This paper proposes an improved A* algorithm optimizing the cost function, search strategy, and node selection. It achieves better multi-objective performance, obstacle avoidance, and memory efficiency, making it ideal for robotic inspection in complex environments.
Abstract: Fifth-generation (5G) heterogeneous networks (HetNets), with ultra-dense small-cell deployments, diverse radio access technologies, and high user mobility, are among the most rapidly evolving infrastructures today. Hence, effective handover management has evolved as an essential requirement for sustaining seamless connectivity, Quality of Service (QoS), and user experience. Conventional handover optimization techniques are generally limited because of high latency, excessive signaling overhead, frequent handover failures, the ping-pong effect, and poor decision-making in a highly dynamic network environment. Recent advancements in quantum computing have opened new opportunities to address these challenges by offering powerful computational and optimization techniques. This survey provides a comprehensive overview of quantum computing and quantuminspired algorithms for handover optimization in 5G HetNets. In particular, it discusses the principles and methodologies of prominent approaches, such as quantum annealing, the Grover search algorithm, Quantum Genetic Algorithms (QGA), Quantum Particle Swarm Optimization (QPSO), quantum- inspired evolutionary algorithms, and Quantum Machine Learning (QML). It further examines their potential to address multidimensional handover optimization problems involving signal quality, user mobility, network load balancing, throughput, latency, energy efficiency, handover failure rate, and QoS requirements. A comparative analysis of the existing studies is provided to evaluate algorithm performance, computational complexity, and scalability for practical deployment in real-world 5G environments. The survey also emphasizes the integration of quantum computing with Artificial Intelligence (AI) and Machine Learning (ML) frameworks for intelligent mobility management. Finally, it highlights some significant research challenges, implementation issues, hardware limitations, and future directions for exploiting quantum technologies in 5G/6G mobility management systems.
Objective: During black-start restoration, grid-forming photovoltaic and battery energy storage systems are subject to weak grid support, frequent operating-condition transitions, and uncertain disturbances, which may lead to frequency deviations, active-power oscillations, and prolonged recovery. To address these challenges, this study develops an adaptive virtual synchronous generator control strategy based on the soft actor-critic algorithm to improve the coordination between transient stability and restoration speed. Method: The online tuning of virtual inertia and damping coefficients is formulated as a continuousaction reinforcement learning problem. Parameter safety bounds, an action-smoothing constraint, and a restoration-time-related reward term are introduced to improve control continuity and balance oscillation suppression with recovery speed. Result: Comparative simulations under load pickup, load shedding, and slave-unit paralleling conditions show that the proposed improved SAC-based strategy achieves smaller frequency deviation, lower active-power overshoot, and shorter recovery time than the fixed-parameter VSG, DDPG, and TD3 methods. Discussion: The results suggest that adaptive tuning of VSG parameters can help balance disturbance suppression and recovery speed in grid-forming photovoltaic and battery energy storage systems during black-start restoration. Since the present validation is based on simulation studies, further investigations considering practical uncertainties, hardware constraints, and more complex operating conditions are needed before engineering application. Conclusion: The simulation results indicate that the proposed strategy can improve the transient regulation and restoration performance of grid-forming photovoltaic and battery energy storage systems during black-start restoration.
Objectives: Traditional quality inspection methods in intelligent manufacturing often suffer from low efficiency, limited accuracy, and poor generalization to real manufacturing environments. This study aims to develop an intelligent quality inspection method with high accuracy and robust generalization capability. Methods: A deep learning model that integrates Convolutional Neural Networks (CNNs) with selfsupervised learning is proposed. CNNs are used to extract key visual features from manufacturing processes. Self-supervised learning is applied for pretraining on unlabeled data to capture latent process representations. In addition, a multi-layer architecture is designed to jointly model spatial and temporal features. An ensemble learning strategy is further introduced to fuse multimodal information and enhance overall detection performance. Results: Experiments were conducted on three quality inspection tasks: printed circuit board solder joints, ceramic substrates, and flexible packaging. The proposed method achieved detection accuracies of 98.1%, 97.3%, and 95.8%, respectively. At a mean average precision (mAP) of 0.5, the convergence speed improved by an average of 33.6%. The inference latency on the same hardware platform was approximately 22 ms. After INT8 quantization, the performance retention reached 98.9%. Discussion: The results indicate that self-supervised pretraining and multi-level feature fusion effectively improve feature representation in complex manufacturing scenarios. At the same time, computational cost is reduced without sacrificing detection accuracy. This enhances the practicality of the method in real industrial environments. Conclusions: The proposed quality inspection method combining CNNs and self-supervised learning outperforms ResNet50 and traditional detection algorithms in terms of accuracy, stability, and computational efficiency. It provides a practical and reliable solution for intelligent manufacturing quality inspection.
Objective: This study aims to quantitatively assess the effectiveness of Vietnam’s energy communication and labeling program for air conditioners by estimating electricity savings, emission reductions, and broader system-level impacts during the 2021-2025 period. Method: The research develops a transparent analytical framework that integrates market data on air conditioner sales and capacity distribution, energy performance characteristics by labeling categories, and seasonal usage patterns. Electricity consumption is calculated under two scenarios: with energy labeling and without energy labeling. The resulting electricity savings are then converted into CO2 emission reductions using national grid emission factors. Results: The analysis indicates that the energy labeling program has delivered consistent and increasing electricity savings over time. During the 2021-2025 period, the average electricity-saving rate is estimated at approximately 15.1%, corresponding to a cumulative saving of about 2.25 billion kWh. These reductions imply a measurable decrease in electricity demand associated with residential air conditioning Discussion: These outcomes suggest that information-based policy instruments can play a meaningful role in influencing energy consumption patterns in the residential sector. By improving transparency regarding appliance performance, energy labeling helps reduce information asymmetry and supports more energy-conscious purchasing behavior. The observed savings are also consistent with broader evidence that demand-side efficiency measures can moderate electricity demand growth and alleviate system stress in rapidly developing economies. Conclusion: Overall, the study provides empirical evidence that energy labeling for air conditioners represents an effective and scalable demand-side policy tool in Vietnam. The findings support the continued use and strengthening of labeling policies, including tighter minimum energy performance standards, periodic revision of efficiency thresholds, and closer integration with complementary demand-side management and public communication strategies.
