Personalized Federated Learning (PFL) has been widely adopted in Mobile Edge Computing (MEC) to enable tailored services without private user data ever leaving the devices. Previous efforts have illustrated that the weighted aggregation determined by the number of samples on clients will hurt the convergence of PFL. Furthermore, the over-personalization of PFL may result in model overfitting and lack of generalization capabilities. Therefore, in this paper, we propose novel quality/discrepancy-aware Multi-task Personalized Federated Learning (MPFL) for MEC to tackle these issues. Specifically, a number of personalized learning objectives of different clients in PFL are considered as multiple tasks. Both local private Batch Normalization (BN) and global shared BN layers are introduced into PFL as the specialized experts to better balance the personalization and generalization capabilities of the local models. Furthermore, the statistical discrepancies between such two BN layers and the model quality of clients are jointly taken into account to design aggregation weights for quick model aggregation with better performance gains. Extensive experiments conducted on three well-known datasets demonstrate that the proposed framework outperforms the state-of-the-art benchmarks in terms of convergence speed and learning accuracy.

The research on human-following robot is important for practical applications. It is a hot field of human–machine technology. This article proposes an adaptive recurrent proportional integral differential (PID) control algorithm with self-tuning filter based on vision to address the issue of insufficient recognition accuracy of specific following targets in the presence of occlusion, multiple people, or deformation. It also aims to further improve the control accuracy and immunity of a human-following robot. First, a depth camera-based red green blue (RGB) picture and a depth image are acquired. The person reidentification algorithm and the YOLOv8 algorithm are used to detect and track the targets. The spatial position information of the targets is calculated by the depth image. Additionally, the orientation proportional differential (PD) controller and the speed proportional integral (PI) controller are built. Its foundation is the discrepancy between the relative posture of the user and the robot. In order to minimize sensor data fluctuations and lessen the negative impacts of relative positional instability, a self-tuning filter is developed. To remember the relative postures between the robot and the user in the history window, an adaptive recurrent mechanism is suggested. The controller has the ability to output the control quantity in an adaptive manner based on the current system state. Finally, experiments are conducted to verify the reliability of the proposed method. The experimental findings demonstrate that the visual pedestrian tracking algorithm proposed in this article is highly adaptable. Compared to the traditional PID, fractional-order PID, and virtual spring model, our method demonstrates significant enhancements, reducing the average distance error by 64.29%, 57.14%, and 60.52% in steering scenarios, and by 42.86%, 40.00%, and 40.00% in straight-ahead scenarios, respectively.

The error compensation and suppression effects of traditional ultrawideband (UWB) ranging in non Line of Sight (NLOS) environments are limited. The contribution of specific occlusions to UWB spatial perception in NLOS is ignored. To achieve comprehensive sensing of spatial information by UWB, we initially analyze the channel impulse response (CIR) and the underlying parameters of registers during UWB communication. By comparing the difference between NLOS and line-of-sight (LOS) environments for each parameter on a continuous time series, a fast discriminative method for UWB environment conversion is proposed. Further, combining the ensemble learning XGBoost classifier, an efficient NLOS occlusion recognition method is proposed. At the same time, an algorithm optimization based on a discrete degree threshold is designed. It is based on loss function probability matrix-weighted predictive labeling. The prediction matrix of the loss function of the XGBoost algorithm is used as label weights. The UWB prediction labels of continuous time series are weighted, which mitigates the effect of low-probability data on the overall prediction results. Finally, UWB spatial sensing experiments are carried out to verify the reliability of the proposed method. The experimental results show that the mutation in the parameter profile can effectively perceive the LOS/NLOS transition. The recognition accuracy of the proposed occlusion recognition method in conditions of human, metal, and wall occlusion is 94.44%, 92.00%, and 95.87%, respectively. In contrast to the origin method, the suggested algorithm’s average recognition accuracy has increased by 16.71%. Its precise recognition accuracy makes UWB spatial sensing more effective.

Many protocols in wireless sensor networks use packet delivery ratio (PDR) as a metric to select the best route, transmission rate or power. PDR is normally estimated either by counting the number of received hello/data messages in a small period of time, i.e., less than 1 second, or by taking the history of PDR into account. The first method is accurate but requires many packets to be sent, which costs too much energy. The second one is energy efficient, but fails to achieve good accuracy. Therefore in this paper we propose a novel estimation method which takes advantage of receiving signal strength. We show with extensive experimental results that the proposed method is 25% more accurate than the second estimation method, while being simple and energy efficient at the same time.

A personal network (PN) is a person-centric network, which creates a distributed environment of a person and provides access to personal resources and services regardless the location of the person. An extra functionality added to PNs can enable different persons to cooperate and to form a group-oriented network called a Federation of Personal Networks (Fednet). Fednet is a temporal, ad hoc, opportunity/purpose driven, secure group-oriented network where the users may be the producers and consumers of the services/content/resources. In this paper, we introduce Fednets and their architecture, explain the lifecycle and the access control process.

Accurate and fast wireless link quality assessment (LQA) for wireless channels would bring huge benefits for mobile multi-hop and multi-rate wireless ad hoc and sensor networks in the form of improved end-to-end performance. In this paper, we propose a novel LQA method based on cross-layer information. The method is implemented in a real test-bed, which is based on IEEE 802.11b/g, and achieved a significant LQA improvement up to 50% in mobile scenarios without introducing overhead. The effectiveness of accurate and fast LQA is demonstrated by feeding it into the routing layer to enable the route decisions to adapt faster during changing situations, especially in mobile scenarios. The experiment results show that the proposed LQA method can lead to faster and smarter routing decisions and higher end-to-end throughput compared to traditional methods, which only use hello packets to determine the link quality in mobile scenario.

Today, many wireless devices are mobile and battery powered, which means that the service duration is highly dependent on energy consumptions and battery capacities. Based on the fact that battery capacity is still limited, energy saving is an important issue in wireless communication. Many methods have been proposed to save energy in wireless communication, such as power-aware routing and sleep mode. However, another important approach, transmission power adaptation, is not yet well discussed. In this paper, we propose a mechanism that selects the appropriate transmission power level to transmit packets instead of using a fixed default power level for all packets, which is common in IEEE 802.11. Lower transmission power levels may result in more retransmissions, but still consumes less energy. In total, energy can be saved and interference can be reduced by selecting a good transmission power level. We propose a two phase power saving mechanism and implemented it in a real test-bed. To evaluate the performance of our mechanism, we used an indoor environment to carry out measurements. The measurement results indicate that a significant amount of energy can be saved for the transmitter in most scenarios compared to using a fixed transmission power level for all packets.

QoS-sensitive applications transmitted over wireless links require precise knowledge of the wireless environment. However, the dynamic nature of wireless channel, together with its different configurations and types, makes so called link quality estimation (LQE) a difficult task. This paper looks into the design of an accurate and fast LQE method for a multi data rate environment. We investigate the impact of various conditions on the LQE accuracy. In result, two different link quality estimation sources, i.e., based on hello packet delivery ratio and signal strength, are measured and their performances are compared. We find that these two methods are not always accurate. As an improvement we propose an adaptive LQE method that chooses different LQE indicators depending on the wireless environment. The performance of the proposed method is verified via an extensive measurement campaign.

With more and more consumer devices becoming capable of wireless communication, the need for good quality ad hoc networking becomes more imminent. Self-configuring and self-adapting multi-hop networking will provide an excellent solution for wireless personal area networks (WPAN) as well as wireless home networks. In such networks, scalability is rarely believed to be a real issue. Instead, the issues are related to the end-to-end quality due to wireless links with constantly and rapidly changing quality. Hence, the network layer not only needs to adapt accurately but also in time. This paper looks into the design of accurate and fast wireless link quality assessment (LQA) options. We investigate the possibility to use cross layer information to improve this process which can be used by the routing layer to adapt faster and more accurately to changing situations. Experiments are done and the result shows the importance of adapting the new LQA options.

Next-generation communications networks will comprise third-generation (3G) cellular and multi-hop ad hoc networks, in particular, the Universal Mobile Telecommunications System (UMTS) and the IEEE 802.11 ad hoc networks. The paper has evaluated the performance of transmission control protocol (TCP) over an integrated UMTS and IEEE 802.11 multi-hop ad hoc network for bulk data transfers. The IEEE 802.11b was used. The TCP performance was evaluated for four dedicated UMTS radio channels with bit rates ranging from 64 kb/s to 2 Mb/s with varying block error rates. The throughput simulation results show that the UMTS radio channel is the performance bottleneck for bit rates up to 128 kb/s. However, for a 2 Mb/s channel, the performance is solely constrained by the IEEE 802.11b multi-hop ad hoc network. We have showed that the performance can be improved by using a larger TCP packet size