Search

The sixth-generation (6G) wireless system has been perceived to be the technology to connect everything. This will generate a huge volume of data traffic, resulting in severe spectrum shortage and system latency. To address this issue, data compression is considered to be indispensable for 6G to achieve efficient data transmissions, increase spectrum efficiency, and reduce system latency. Consequently, data compression technologies based on machine learning and deep learning, commonly known as learning-based data compressions, have received intensive attention lately. Taking the industrial IoT (IIoT) as a use case, this article attempts to explore the latest research progress on learning-based data compressions toward efficient data transmissions. Specifically, we first propose a novel learning-based data compression framework for edge-cloud collaborative IIoT. Then, we summarize the learning-based data transmission methods which are involved with various layers of the proposed edge-cloud collaborative framework. Moreover, we conduct a case study to show that our learning-based data transmission methods can effectively reduce the volume of the transmitted data. Finally, we highlight several promising future research directions on the learning-based data compression, such as robustness and instability of deep models, deep model optimization, and future deployment strategies for 6G enabled IIoT. [ABSTRACT FROM AUTHOR]

In this paper, we study the industrial iot applications for AI edge computing technology. We discuss about the edge AI technology that is considered the combination of AI with edge computing and provide an overview of edge AI applications for IIoT networks, where the following three challenges are important to address: a) personalization, b) responsiveness and c) privacy preserving. To this end, we propose a federated active transfer learning (FATL) model, which through training and testing is able to address those open challenges. Details about the training and testing of the proposed FATL global model are given including the corresponding simulation setup. Here it concludes with a discussion and comparison of the simulation results with existing AI edge training solutions, where it provides useful insights about the proposed FALT model. [ABSTRACT FROM AUTHOR]

In this paper, we present an oneM2M-compliant system including an artificial intelligence of things (AIoT) edge device whose principal function is to estimate human poses by using two PoseNet models built on MobileNet v1 and ResNet-50 backbone architectures. Although MobileNet v1 is generally known to be much faster but less accurate than ResNet50, it is necessary to analyze the performances of whole PoseNet models carefully and select one of them suitable for the AIoT edge device. For this reason, we first investigate the computational complexity of the models about their neural network layers and parameters and then compare their performances in terms of inference time, frame rate and pose score. We found that the model with MobileNet v1 could estimate human poses very fast without significant pose score degradation when compared with the model with ResNet-50 through some experimental results. In addition, the model with MobileNet v1 could smoothly handle video clips with 480 × 360 and 640 × 480 resolutions as inputs except 1280 × 720 resolution. Finally, we implemented the overall oneM2M-compliant system with an AIoT edge device containing the model with MobileNet v1 and a floating population monitoring server, which is applicable to commercial area analysis in a smart city and we verified its operation and feasibility by periodically displaying all the data of temperature, humidity, illuminance, and floating population estimation from the AIoT edge device on a map in a web browser. [ABSTRACT FROM AUTHOR]

In this article, we study edge artificial intelligence (AI) for industrial Internet of Things (IIoT) applications. We discuss edge AI technology, which is considered the combination of AI with edge computing, and provide an overview of edge AI applications for IIoT networks, where the following three challenges are important to address: 1) personalization, 2) responsiveness, and 3) privacy preservation. To this end, we propose a federated active transfer learning (FATL) model, which through training and testing is able to address those open challenges. Details about the training and testing of the proposed FATL global model are given, including the corresponding simulation setup. This work concludes with a discussion and comparison of the obtained simulation results with existing edge AI training solutions, which provide useful insights about the proposed FATL model. The simulation results highlight how the FATL global model can efficiently address the open challenges of edge AI for future IIoT applications. [ABSTRACT FROM AUTHOR]

In this paper, an adaptive edge service placement mechanism based on online learning and a predictive edge service migration method based on factor graph model are proposed to solve the edge computing service placement problem from the edge computing dimension. First, the time series of the development of online chaotic public opinion is a platform for vectorized collection of keyword index trends using the theory of chaotic phase space reconstruction. Secondly, it is necessary to use the main index method to judge whether the time series has the chaotic characteristics of the network public opinion data. The simulation results show that network public opinion is the development characteristic of chaotic time series. Finally, the prediction model is improved by using complex network topology. Through the simulation experiment of network public opinion and chaotic time series, the results show that the improved model has the advantages of accuracy, rapidity, and self-adaptability and can be applied to other fields. [ABSTRACT FROM AUTHOR]