Building adaptive algorithms for indoor heating control has been a well-established area of research. These methods, favored for goals such as cost saving and ultimately enhancing occupants' thermal comfort experience, still pose significant challenges. Controllers, such as the Model Predictive Controller (MPC), require an elaborate and precise system model, which is usually expensive to develop. Data-driven models can assist with the identification of the Grey-box system thermal model without extensive prior knowledge. This study presents a comparative analysis and investigation of four analytical models, namely Non-linear Least Squares (NLS), Unscented Kalman Filter (UKF), and Physics-Informed Neural Networks (PINN), and comparison against the conventional Continuous Stochastic Thermal modeling (CSTM) method to demonstrate their applicability and precision. The analysis are conducted on short- and long-term validation of two real-world experimental case studies. The results indicate that UKF has shown better performance for parameter estimation of the building. It reached an RMSE and MAPE of 0.66∘C and 2.1% on the short-validation and 0.49∘C and 1.7% for long-validation datasets. It was followed by CSTM, NLS, and PINN algorithms. While the models generally perform well on the short validation dataset, the low R2 scores for the long-validation dataset indicate less stability across all models on the long-validation dataset.

The teaching and learning activities (TLAs), such as a written report in the form of a research assignment, an introductory lecture, presentations, and discussion, help students practice and engage with their projects while developing crucial skills, including teamwork and collaboration. These TLAs support students in reaching higher levels of learning, beyond the basic remember level in Bloom's taxonomy or the uni-structural level in the SOLO taxonomy. The aim of this study is to develop the research components for the capstone project in computer science, engineering and data science courses, which are commonly offered in bachelor's or master's programs in computer science, engineering, and data science. It also seeks to promote a deep learning approach while taking into account the diversity of students' backgrounds and their prior experiences in computer science, engineering or data science. The structure presented in this study can help instructors responsible for capstone projects or similar courses in computer science, engineering, or data science to implement a project-based learning approach. By doing so, students can gain preliminary research experience during the course, better preparing them for their bachelor's or master's thesis.

As malware continues to evolve in complexity and diversity, traditional single-modal detection approaches demonstrate limited discriminative capacity and struggle to generalize to variants. Although existing approaches attempt to capture program semantics and logic, current designs still struggle to exploit complementary information across heterogeneous modalities. Mechanisms for quantifying inter-modal conflicts, modeling uncertainty, and moving beyond perturbation-prone surface features remain underdeveloped. To address these limitations, we propose DSMalConv, a multi-modal malware detector that fuses byte-level Grayscale Images with Attributed Control Flow Graphs (ACFGs). Convolutional Neural Network (CNN) modules extract visual patterns, while Graph Neural Network (GNN) modules model execution flow dependencies, yielding complementary representations. We further integrate Dempster-Shafer (DS) evidence theory to identify uncertain samples, incorporate uncertainty into the training loss function, and provide test-time feedback to facilitate safer automated decisions. Experimental evaluations on cross-architecture datasets demonstrate that DSMalConv achieves an average accuracy of 99.76% and F1 of 99.75%, outperforming strong baselines. Analysis reveals that shallow features and multi-modal methods achieve only 94.21% to 97.65% accuracy due to insufficient representations and unresolved inter-modal conflicts, leading to misclassifications. DSMalConv overcomes these limitations through dual-modal fusion and an uncertainty-aware decision mechanism, achieving robust detection under high-conflict conditions. Robustness evaluations against adversarial attacks targeting both visual and structural representations demonstrate that DSMalConv exhibits superior resistance through credible fusion of multi-modal features. The ablation studies further validate the effectiveness of the multi-modal fusion strategy.

Due to the breakthroughs in Wireless Power Transmission (WPT) technology, achieving long-term stable Wireless Rechargeable Sensor Networks (WRSNs) is now possible. Traditional periodic charging schemes cannot adequately adapt to dynamically changing environments or sensor energy fluctuations and demands. In most existing studies on multi-MCs (Mobile Chargers) on-demand charging schemes, few can ensure that multiple MCs can cooperate to replenish energy to all nodes without interfering with each other's charging tasks. To address these issues, we propose a Multi-MCs Cooperative Charging Scheduling (MMCC) scheme. Firstly, we introduce a node group construction algorithm to balance charging tasks. Then, a reasonable charging request threshold is set for nodes based on their locations and energy consumption rates. Next, three types of MCs are adopted in each group to charge nodes in a distributed and collaborative manner. At last, we propose a node locking strategy to prevent "charging conflicts" between MCs and "overcharging" of nodes. Simulation results show that MMCC performs well in terms of node survival rate and average waiting time for charging after nodes send out charging requests.

With the increasing availability and affordability of building-integrated Heat Pumps (HPs), the number of heat pumps installed in residential buildings has risen significantly in recent years. When coupled with conventional District Heating (DH) systems in a hybrid setting, HPs provide higher energy reliability and cost-effective solutions for domestic heating. The operation of such systems, however, requires a sophisticated control system that simultaneously considers the dynamics of energy pricing and building energy needs. In this paper, we propose a nonlinear economic model predictive control to determine the optimal share for a hybrid DH-HP heating system. A resistor-capacitor thermal building model is utilized to capture the system dynamics. The results indicate that the proposed controller in the hybrid DH-HP system has a cost saving between 29% and 57% compared to the baseline scenario.

The concept of smart building is based on the deployment of Internet of Things (IoT) technologies to develop various building applications and services. Aided by the proliferation of smart devices, research in building automation has grown significantly. Nevertheless, these smart devices are integrated with sensors that can collect and share sensitive data and private information related to the building occupants, exposing them to a variety of privacy threats. Although research efforts to promote the development of privacy-aware solutions for smart buildings have been on the rise, a comprehensive review that summarizes these studies is lacking in the literature. This paper provides an extensive review of the studies related to sensing in smart buildings. It highlights privacy issues connected to sensing in smart buildings, provides mitigation strategies that can be deployed to minimize occupants’ privacy invasions, and discusses future research directions towards realising privacy-aware smart buildings. To fulfill the aim of this study, five research questions are formulated, which enable systematic navigation through existing studies related to the topic. These research questions are directed to providing answers to privacy related to data leakage, privacy connected to sensor types, privacy related to different applications, privacy concerns with sensor deployment locations and building types, privacy issues with data processing methods, and to highlight mitigation strategies for reducing privacy invasion. It further discusses the technical approaches, general principles, and design choices for privacy-aware applications which are relevant for guiding relevant stakeholders.

Artificial intelligence (AI) has become a cornerstone of modern consumer electronics (CE), seamlessly integrating into domains, such as FinTech, smart homes, autonomous driving, and information security. By processing diverse data types, such as voice, images, videos, and wireless radio-frequency signals, AI has enhanced feature extraction, prediction, and recognition, making everyday life more convenient. However, as reliance on AI systems grows, adversarial examples (ADVs) pose a growing threat, exploiting vulnerabilities in AI models with imperceptible perturbations. These deceptive attacks have significant implications for CE systems, where security lapses can result in far-reaching societal and economic consequences.

Unmanned aerial vehicles (UAVs) are increasingly utilized in smart city applications, particularly for urban surveillance. UAVs can be provisioned as mobile surveillance to avoid various difficulties in ground operations and reduce extensive labor cost. However, their limited energy capacity restricts flight time and coverage, making it difficult to build a large-scale, long-term city-wide monitoring network. To address this problem, an ubiquitous public transportation network is introduced for UAVs to periodically recharge by landing on public transportation buses. We propose a novel public transportation-assisted UAV scheduling framework that leverages the existing bus network to enable recharging of UAVs. Two intertwined sub-problems are addressed: the UAV trajectory planning and the surveillance task offloading problems. The trajectory planning is modeled as a Traveling Salesman Problem (TSP) and it is solved via the Lin-Kernighan heuristic (LKH), decomposing the bus station network into sub-graphs for efficient routing. For task offloading, a time-slot-based scheduling method is proposed that dynamically assigns UAVs to monitor points of interest (PoIs) while ensuring energy constraints and full coverage. Experimental results demonstrate that the proposal outperforms baseline algorithms, achieving a 1.72%-3.46% average extension in system lifetime compared to state-of-the-art baselines, while maintaining computational efficiency (average runtime: 277.57 ms). The robustness of the proposal is further validated across diverse testing instances with various parameter settings.

Currently, utilizing Mobile Vehicles (MVs) equipped with both wireless charging and data transmission capabilities to recharge nodes and collect their data in a 'parallel' manner has become an effective approach to enhancing the efficiency of Wireless Rechargeable Sensor Networks (WRSNs). However, how to address the varying types and frequencies of service requests from nodes while minimizing the data loss ratio remains a critical issue that needs to be solved. To this end, this paper proposes an energy Replenishment and data Collection strategy based on Minimizing the data Loss ratio (RCML). First, the theoretical upper bound of service duration per round for MV was calculated, and reasonable service request thresholds were set for nodes accordingly. Then, an objective function was constructed with the primary and secondary goals of reducing data loss and minimizing travel duration of MV, respectively. Based on this, a Service Queue Generation algorithm (SQG) which utilizes the simulated annealing was proposed. To address the issue of a large number of requests, a Node Selection Strategy (NSS) was also proposed to prioritize the nodes at higher risk of data loss. Finally, an Idle-Time Service (ITS) strategy was adopted to further improve the overall service efficiency of MV. Simulation results show that RCML demonstrates significant advantages in terms of data loss ratio as well as energy consumption of MV compared to typical methods such as MPF and PMCDC.

Holographic video communications epitomize a paradigm shift within the domain of visual communications. Arising from the amalgamation of advancements in optics, computational algorithms, and display technologies, holography transcends the confines of conventional communication modalities. The capacity to construct three-dimensional, lifelike depictions of information heralds unprecedented frontiers in telepresence, data visualization, and user interaction. In recent years, concerted efforts within relevant domains have driven remarkable advancements in holographic displays, imaging, and telepresence, establishing holography as a transformative technology across diverse sectors. From healthcare and education to entertainment and collaborative workspaces, holographic video communications fundamentally reshape our perceptual and interactive engagement with information, meriting acknowledgment as a significant component of the 6G Vision outlined in International Mobile Telecommunications (IMT) towards 2030.