Precise measurements from sensors are crucial, but data is usually collected from low-cost, low-tech systems, which are often inaccurate. Thus, they require further calibrations. To that end, we first identify three requirements for effective calibration under practical low-tech sensor conditions. Based on the requirements, we develop a model called TESLA, Transformer for effective sensor calibration utilizing logarithmic-binned attention. TESLA uses a high-performance deep learning model, Transformers, to calibrate and capture non-linear components. At its core, it employs logarithmic binning, to minimize attention complexity. TESLA achieves consistent real-time calibration, even with longer sequences and finer-grained time series in hardware-constrained systems. Experiments show that TESLA outperforms existing novel deep learning and newly crafted linear models in accuracy, calibration speed, and energy efficiency.

Bees play a crucial role in human food production and ecosystem maintenance. However, they face a global crisis characterized by significant population decline, including colony collapse disorder, which threatens their survival. Therefore, the development of precision beekeeping Internet of Things (IoT) systems is pivotal for enhancing bee colony strength. This study analyzes and categorizes the research focusing on the colony strength into two main areas based on colony activity and threat detection. Research on colony activity aids beekeepers to effectively manage their hives by analyzing behaviors, such as internal status, swarming, and traffic. In threat detection, research focuses on identifying critical predators, such as hornets and varroa destructor, that significantly affect colony survival. This study conducts a comprehensive literature review, thoroughly presenting findings on the evolution and diversification of methodologies aimed at enhancing colony strength. By delving into various innovative approaches and assessing their effectiveness, this review highlights key developments that have significantly contributed to improving the resilience of bee populations against emerging threats. Furthermore, this study identifies the limitations of current research and proposes future research directions focused on enhancing the accuracy of bee behavior analysis and threat detection to improve colony strength and productivity.

The collection of accurate and noise-free data is a crucial part of Internet of Things (IoT)-controlled environments. However, the data collected from various sensors in daily life often suffer from inaccuracies. Additionally, IoT-controlled devices with low-cost sensors lack sufficient hardware resources to employ conventional deep learning models. To overcome this limitation, we propose sensors for daily life (SenDaL), the first framework that utilizes neural networks for calibrating low-cost sensors. SenDaL introduces novel training and inference processes that enable it to achieve accuracy comparable to deep learning models while simultaneously preserving latency and energy consumption similar to linear models. SenDaL is first trained in a bottom-up manner, making decisions based on calibration results from both linear and deep learning models. Once both models are trained, SenDaL makes independent decisions through a top-down inference process, ensuring accuracy and inference speed. Furthermore, SenDaL can select the optimal deep learning model according to the resources of the IoT devices because it is compatible with various deep learning models, such as long short-term memory-based and Transformer-based models. We have verified that SenDaL outperforms existing deep learning models in terms of accuracy, latency, and energy efficiency through experiments conducted in different IoT environments and real-life scenarios.

The Internet of Things (IoT) connects a wide range of entities and can be applied to various types of environments. In addition, IoT environments can be dynamically changed at runtime; thus, IoT systems can be deployed in various environments. To support stable operation, IoT systems must adapt to dynamic environmental changes. The self-adaptive software aims to adjust various artifacts or attributes of software to adapt the detected context by itself, and various studies have applied self-adaptive methods in IoT-related research. In this study, we proposed a self-adaptive software framework with master–slave architecture-based finite-state machine modeling. In addition, model checking is applied, which is a formal method to verify IoT systems at runtime, and a cache-based mechanism is applied to reduce the computational time required for verification. To demonstrate the efficiency of the proposed framework, an empirical evaluation was performed with several model-checking tools (RINGA, NuSMV, nuXmv, and CadenceSMV), and the results showed the efficiency of the proposed framework with the cache-based mechanism. In addition, an example application was investigated with smart greenhouse scenarios, and the application was implemented on Android and Arduino. The application was operated in physical environments, and the results showed the practical usability of the proposed framework with verification at runtime.

Currently, software systems are operated in a dynamic and uncertain environment, making it difficult to predict the operating environment. Particularly, the Internet of Things (IoT) interconnects several entities, users, and information resources with services. Therefore, IoT systems can dynamically create various environments at runtime. Consequently, to support the dynamic IoT environment, an efficient verification method is required for IoT systems. Model checking is one of the formal methods for verification of concurrent systems, which is applied in several software fields. However, model checking has a chronic problem (i.e., state explosion) that obstructs the verification at runtime. To overcome this limitation, one possible approach is to abstract the system design as expression forms and then perform verification by solving the expressions. To apply this approach, the abstraction and verification processes need to be performed at a reasonable time. In this article, we focused on developing an efficient model checking method during the IoT system runtime verification. A cache mechanism is proposed that reduces the computational time for abstraction and verification, which was validated through experiments. Additionally, the comparison of other model checking tools (such as RINGA, CadenceSMV, NuSMV, and nuXmv) reveals that the proposed approach is more efficient at runtime.

The Internet of Things (IoT) connects several objects within environments that dynamically change, and so requirements may be added and changed at runtime. Therefore, requirements may be satisfied at dynamic change. Self-adaptive software can alter their behavior to satisfy requirements in dynamic environments. In this perspective, the concept of self-adaptive software is suitable for IoT environments. In this study, a self-adaptive framework is proposed for decision making in IoT environments at runtime. The framework includes finite-state machine model designs and game theoretic decision-making methods to extract efficient strategies. The framework is implemented as a prototype, and experiments are performed to evaluate runtime performance. The results demonstrate that the proposed framework can be applied to IoT environments at runtime.

Twitter is a microblogging website, which has different characteristics from any other social networking service (SNS) in that it has one-directional relationships between users with short posts of less than 140 characters. These characteristics make Twitter not only a social network but also a news media. In addition, Twitter posts have been used and analyzed in various fields such as marketing, prediction of presidential elections, and requirement analysis. With an increase in Twitter usage, we need a more effective method to analyze Twitter content. In this paper, we propose a method for content analysis based on the influence of Twitter content. For measuring Twitter influence, we use the number of followers of the content author, retweet count, and currency of time. We perform experiments to compare the proposed method, frequency, numerical statistics, user influence, and sentiment score. The results show that the proposed method is slightly better than the other methods. In addition, we discuss Twitter characteristics and a method for an effective analysis of Twitter content.