Special issue on smart interactions in cyber-physical systems: Humans, agents, robots, machines, and sensors

Abstract

In recent years, there has been increasing interaction between humans and non-human systems as we move further beyond the industrial age, the information age, and as we move into the fourth-generation society. The ability to distinguish between human and non-human capabilities has become more difficult to discern. Given this, it is common that cyber-physical systems (CPSs) are rapidly integrated with human functionality, and humans have become increasingly dependent on CPSs to perform their daily routines. The constant indicators of a future where human and non-human CPSs relationships consistently interact and where they allow each other to navigate through a set of non-trivial goals is an interesting and rich area of research, discovery, and practical work area. The evidence of convergence has rapidly gained clarity, demonstrating that we can use complex combinations of sensors, artificial intelligence, and data to augment human life and knowledge. To expand the knowledge in this area, we should explain how to model, design, validate, implement, and experiment with these complex systems of interaction, communication, and networking, which will be developed and explored in this special issue. This special issue will include ideas of the future that are relevant for understanding, discerning, and developing the relationship between humans and non-human CPSs as well as the practical nature of systems that facilitate the integration between humans, agents, robots, machines, and sensors (HARMS). Contributions that demonstrate the integration of HARMS using practical experimental results were invited for this issue. Papers that show the design, models, or techniques of advancement were selected in the wider context of large, complex systems, including those involving multiple, heterogeneous actors. The first paper “Deep Compression of Convolutional Neural Networks with Low-Rank Approximation,” by Marcella Astrid and Seung-Ik Lee, captures the application of deep neural networks (DNNs) to connect the world with cyber-physical systems (CPSs), which have attracted much attention. However, DNNs require a large amount of memory and computational cost, which hinders their use in the relatively low-end smart devices that are widely used in CPSs. In this paper, the authors aim to determine whether DNNs can be efficiently deployed and operated in low-end smart devices. To do this, they develop a method to reduce the memory requirement of DNNs and to increase the inference speed, while maintaining the performance (for example, accuracy) close to the original level. The parameters of DNNs are decomposed using a hybrid of canonical polyadic–singular value decomposition, and are approximated using a tensor power method, and they are fine-tuned by performing iterative one-shot hybrid fine-tuning to recover from the decreased accuracy. In this study, they evaluate their method on frequently used networks. The authors also present results from extensive experiments on the effects of several fine-tuning methods, as well as the importance of iterative fine-tuning and decomposition techniques. The authors demonstrate the effectiveness of the proposed method by deploying compressed networks in smartphones. The second paper, “Human-like Sign Language Learning Method with Deep Learning,” by Ki-Baek Lee and others proposes a human-like sign language learning method with a deep learning technique. Inspired by the fact that humans can learn sign language from just a set of pictures in a book, in the proposed method, the input data are pre-processed into an image. In addition, the network is partially pre-trained to imitate the preliminarily obtained knowledge of humans. The learning process is implemented with a well-known network, that is, a convolutional neural network. Twelve sign actions are learned in ten scenarios and can be recognized with an accuracy of 99% in an environment involving low-cost equipment and limited data. The results demonstrate th