Artificial intelligence (AI) and machine learning (ML) are widely employed to make the solutions more accurate and autonomous in many smart and intelligent applications in the Internet of Things (IoT). In these IoT applications, the performance and accuracy of AI/ML models are the main concerns; however, the transparency, interpretability, and responsibility of the models’ decisions are often neglected. Moreover, in AI/ML-supported next-generation IoT applications, there is a need for more reliable, transparent, and explainable systems. In particular, regardless of whether the decisions are simple or complex, how the decision is made, which features affect the decision, and their adoption and interpretation by people or experts are crucial issues. Also, people typically perceive unpredictable or opaque AI outcomes with skepticism, which reduces the adoption and proliferation of IoT applications. To that end, explainable AI (XAI) has emerged as a promising research topic that allows ante-hoc and post-hoc functioning and stages of black-box models to be transparent, understandable, and interpretable. In this article, we provide an in-depth and systematic review of recent studies that use XAI models in the scope of the IoT domain. We classify the studies according to their methodology and application areas. Additionally, we highlight the challenges and open issues and provide promising future directions to lead the researchers in future investigations.
In recent years, Internet of Things (IoT) concept has become a promising research topic in many areas including industry, commerce and education. Smart cities employ IoT based services and applications to create a sustainable urban life. By using information and communication technologies, IoT enables smart cities to make city stakeholders more aware, interactive and efficient. With the increase in number of IoT based smart city applications, the amount of data produced by these applications is increased tremendously. Governments and city stakeholders take early precautions to process these data and predict future effects to ensure sustainable development. In prediction context, deep learning techniques have been used for several forecasting problems in big data. This inspires us to use deep learning methods for prediction of IoT data. Hence, in this paper, a novel deep learning model is proposed for analyzing IoT smart city data. We propose a novel model based on Long Short Term Memory (LSTM) networks to predict future values of air quality in a smart city. The evaluation results of the proposed model are found to be promising and they show that the model can be used in other smart city prediction problems as well.
Internet of Things (IoT) devices generate a vast amount of sensing data. Reliability of this data is a vital issue to ensure IoT service quality. However, IoT data usually suffers from missing and incomplete values due to various reasons such as noise, collision, unstable network communication, equipment failure and manual system closure. Transferring all IoT data to cloud level to solve missing data problem may negatively affect network performance and service quality due to excessive latency, bandwidth limitation, and high communication costs. Therefore, missing data problem should be taken care of as early as possible by offloading tasks such as data prediction or estimation closer to the edge devices in the network. In this paper, we propose a missing data prediction protocol called DeepMDP for IoT systems with unreliable data sources, which can reduce the amount of data transmission and delay in the network significantly. The proposed protocol can work on resource-constrained IoT devices as well as fog and cloud servers. Besides, to evaluate the proposed protocol, we design a real-world testbed architecture called DeepArch consisting of edge, fog, and cloud layers. Under several application scenarios, we evaluate the efficiency of DeepMDP on the DeepArch platform. Experimental results show that the proposed protocol can significantly reduce the amount of data transmission and delay while accurately predicting missing data.
20 June 2019 in Concurrency and Computation: Practice and Experience
This paper presents the development and application of a Digital Twin (DT) model for the optimization of micro-grid operations. With the increasing integration of renewable energy resources (RERs) into power grids, micro-grids are essential for enhancing grid resilience and sustainability. The proposed DT model, enhanced with Deep Reinforcement Learning (DRL), simulates and optimizes key micro-grid functions, such as battery scheduling and load balancing, to improve energy efficiency and reduce operational costs. The model incorporates real-time monitoring, service-oriented simulations, cloud-based deployments, "what-if" analyses, advanced data analytics, and security features to enable comprehensive management of DTs. An optimization scenario was conducted to evaluate the effectiveness of the DT and DRL in improving micro-grid performance. The results demonstrated significant revenue improvements: 81.7% for PPO and 56.12% for SAC compared to the baseline. These findings highlight both the promising potential of DT technology and the critical importance of incorporating DRL techniques into the DTs to improve system performance and resilience.