Activity recognition models based on wearable devices are becoming increasingly popular. However, models that are trained and tested on the same players show a large bias and are not generalizable to previously unseen players. In this paper, we tackle the badminton stroke recognition problem from this perspective, comparing the performance of individual and generalized models based on an accelerometer and a gyroscope, and identifying which components of the solution can maximize the performance of generalized models. First, we describe a simple convolutional neural network trained to classify 7 types of stroke. Second, the model is extended in a hybrid way to identify two additional classes (movement and rest). Third, data augmentation is applied on the training set. Fourth, transfer learning is applied to use data from the test player to fine-tune the generalized model and attempt to reach the performance of an individual model. These models are evaluated on a dataset collected from amateur players, both in a controlled environment and in a match simulation. The results showed a large difference between the performance of individual and generalized models; however, the latter could be improved by increasing the number of players in the training set, by data augmentation, and by transfer learning, highlighting the necessity of larger datasets in this field.
Digital Twin Networks are advanced digital replicas of physical network infrastructures, offering real-time monitoring, analysis, and optimization capabilities. Despite their potential, the absence of a standardized definition and implementation guidelines complicates practical deployment. The existing literature often lacks clarity on tool selection and implementation specifics. In response, this paper aims to address these challenges by providing a complete guide and reference list of essential tools to implement Digital Twin Networks. Following the current research and work-in-progress from the definition initiative, including our own contributions, we propose a structured approach to Digital Twin Network implementation. Our methodology integrates insights from diverse sources to establish a coherent framework for developers and researchers. By synthesizing insights from the literature and practical experience, we define key components and functionalities critical to Digital Twin Network architecture. Additionally, we highlight challenges inherent to Digital Twin Network implementation and offer strategic approaches and mindsets for addressing them. This includes considerations for scalability, interoperability, real-time communication, data modeling, and security, ensuring a holistic approach to building effective Digital Twin Network systems.
This paper presents a performance evaluation of Bluetooth Channel Sounding, a feature introduced in Bluetooth 6.0, tested on commercially available hardware. The evaluation includes a publicly accessible dataset containing raw Channel Sounding measurements annotated with millimeter-accurate ground truth, and metadata of the measurement campaign. Results demonstrate that Channel Sounding is well-suited for its primary envisioned use case, which is secure proximity detection in keyless entry systems. Its accuracy at close range ensures reliable unlocking when the user is truly nearby, while errors at longer distances are less critical, as access would not be granted regardless. Moreover, it also shows promise for indoor localization, offering encouraging initial results for future deployment of localization with Channel Sounding. However, the presence of outliers can significantly affect accuracy and must be accounted for in practical applications. The evaluation underscores the impact of algorithm choice, configuration parameters, and environmental conditions on performance. While the approach is promising, single-antenna setups currently lack the robustness needed for high-precision localization, highlighting the need for more accurate algorithms and system-level improvements. The dataset and analysis provide a foundation for future improvements through advanced signal processing and machine learning techniques.
Bluetooth Mesh (BM) technology is a suitable candidate to realize mesh networks leveraging on a mains-powered backbone. However, the flooding-based technology requires additional management in order to operate efficiently. Consequently, it offers a diverse set of configuration options to control network behavior, including the selection of relays in the backbone that rebroadcast packets further into the network. In the past, Digital Twin technology has been applied to design a Digital Communication Twin (DCT) of a BM network. The DCT (or Digital Twin Network, DTN) can be used to find an optimal network configuration for given application requirements. This includes a relay selection approach that finds a set of relays to assure sufficient path redundancy in the network. However, existing results prove that this approach can be improved in terms of computation time, flexibility and validity of the proposed set of relays. In this paper, we present a reinterpreted version of a genetic algorithm to tackle this issue. The results show it achieves better results on all three improvement targets, compared to the original approach. Furthermore, it showcases the flexibility and adaptability of the DCT whilst adequately improving the relay selection approach.
We present a wireless multi-hop network management system that can capture the current network context to gain continuous insights into network performance and provide reconfiguration suggestions. Ongoing IETF standardization efforts refer to the creation of this type of virtual representation as a Digital Twin Network. The management system presented in this paper is based on Bluetooth Mesh networks and combines simulation, graph-based algorithms and theoretical models to gain insights into network performance and provide per-flow reconfiguration suggestions. Through an advanced testbed, we showcase the efficiency of Digital Twin Network-based configurations compared against simple configurations. The demonstration highlights the system’s ease of use and swift implementation with a plug-and-play approach. This study paves the way for the advancement of network management in wireless communication systems and more efficient traffic scheduling for constrained Internet of Things devices. It underscores the potential of autonomous Digital Twin Network configurations to improve network resilience and performance.