High-Altitude Platform Stations (HAPSs), such as airships and balloons, operate in the stratosphere. By establishing wireless interconnections among multiple HAPS units, a HAPS mesh can be formed to cover a vast terrestrial area. The network can thus provide an ideal platform for broadcasting signals to large and geographically dispersed recipients. A service we call Selected Group Broadcasting (SGB) is for delivering messages to members of specific subscriber groups within these broad audiences. In essence, the routing problem underneath SGB is essentially a Steiner Tree Problem in Graphs (STPG). In this paper, we introduce a novel Cell-Expansion algorithm, a reduction-integration framework that offers a conditionally exact solution for STPG with a preset number of iterations. By leveraging the small-world network characteristics inherent in HAPS meshes, where the network diameter is naturally small, we further propose the Region-Cut-Split algorithm for merging inter-regional networks, along with a Regional-Cell-Inclusion test mechanism. The integration of these algorithms yields a provably optimal solution for SGB routing under HAPS topological conditions. The distributed and parallel nature of the algorithms enables the provision of feasible real-time QoS-assured SGB services for data delivery across HAPS mesh networks. Thorough simulations and verification confirm the effectiveness of our proposed designs in constructing SGB routing over arbitrary HAPS topologies. Our solution consistently outperforms state-of-the-art heuristics in both accuracy and efficiency, while surpassing exact solvers in real-time performance.

The High Altitude Platform Station (HAPS) mesh network can cover a large surface area on Earth, making it an excellent choice for broadcasting signals to a vast number of recipients distributed across extensive regions simultaneously. Different applications may attract different sets of audiences, leading to various groups of recipients subscribing to different content services. We refer to data delivery among these subscribers as Selected Group Broadcasting (SGB). Essentially, SGB routing is a Steiner Tree Problem in Graphs (STPG). In this paper, we propose a Cell-Expansion algorithm that provides a generic exact solution for hop-constrained STPG and functions as a reduction-integration framework. The HAPS mesh follows the concept of a small-world network, characterized by its unique topological properties. Accordingly, we introduce the Region-Cut-Split algorithm for merging inter-regional networks. By combining these algorithms, we achieve an exact solution for SGB routing. The distributed parallel runtimes of our algorithms enable feasible real-time, QoS-assured SGB services for data delivery across a HAPS mesh network. Rigorous simulations and verification confirm the efficacy of our proposed mechanisms for constructing SGB routing on arbitrary random HAPS topologies. In HAPS mesh simulations, our solution consistently outperforms state-of-the-art heuristics in both accuracy and efficiency, and surpasses exact solvers in real-time performance.

There are numerous decentralized and distributed computing applications running across the Internet today. In some applications such as those consensus-related, senders may dispatch duplicated messages, i.e., one-to-many identical messages at the application layer, to reach multiple recipients simultaneously. To effectively characterize the performance of such systems, tractable and accurate traffic models are highly desirable. Although one-to-one traffic modeling, e.g., M/M/1 model, has been extensively studied, traffic characterization for one-to-many connectivity remains underdeveloped. To accurately understand the performance of applications that generate one-to-many traffic, closed-form solutions are always desirable. For example, the hyper-Erlang distribution accurately models the residence time and channel holding time of a user in a wireless cellular network. However, the relationship between the coefficients of the hyper-Erlang distribution and system parameters or control variables is often unclear. In this paper, we find the traffic distribution for one-to-many traffic delay distributions over a Selected Group Broadcasting (SGB) in a wide-area communication system. Given the system parameters of an SGB model, we obtain closed-form formulas for the density functions of both the first and second order statistics – mean and variance – of the latency density functions of a broadcast overlay. Through simulations, we verify that the derived hyper-Erlang and modified gamma processes align well with the mean and variance of the simulated delays, as confirmed by the Kolmogorov-Smirnov test.

With the advent of autonomous intelligent systems, such as humanoid robots, environmental perception should require rapid and accurate real-time 3D scene interpretation. LiDAR sensors are core and accurate distance measuring components, but it is complicated to process the unstructured, sparse, and unevenly distributed nature of LiDAR point cloud data while meeting the real-time object classification needs. To address the limitations of current 3D LiDAR-based panoptic segmentation methods, we propose a MambaPan3D design. It is a hybrid architecture that integrates Mamba and Transformer models for efficient and accurate 3D point cloud understanding. Our framework solves two key challenges: 1) geometric ambiguity caused by sparse and irregular LiDAR point cloud distributions, and 2) inefficient long-range dependency modeling in largescale scenes. Specifically, CartPolar-KAN embedding, a novel positional encoding strategy, is introduced to interpret between Cartesian and polar coordinates by adding a Kolmogorov-Arnold network (KAN) with learnable B-spline basis functions. The module dynamically fuses multi-coordinate features to overcome the limitations of fixed Bird's-Eye View (BEV) quantization. Additionally, our Mamba-Transformer Decoder combines the global attention capabilities of the Transformer and the linear computational efficiency of the Mamba state-space model to achieve real-time inference while maintaining the global receptive field. Extensive experiments on SemanticKITTI dataset demonstrated state-of-the-art performance. The panoptic quality (PQ) could reach 63.3 % in complex urban scenes, i.e., 1.3 % higher than the current optimal baseline method. The proposed framework provides a powerful solution for real-time situational awareness in autonomous driving systems, balancing accuracy, efficiency, and scalability. Our MambaPan3D model offers a robust solution for real-time situational awareness in autonomous systems through balancing accuracy, efficiency, and scalability.

Smart patient wristbands are expected to store patient information, monitor real-time vital signs, notify nurses for timely and emergency services, alert patients about safety issues, and assist with daily hospital operations. To better track patient movements, especially those with infectious diseases, these smart loT wristbands are equipped with low-cost, MRI-safe Bluetooth Low Energy (BLE) transceivers. In this paper, we propose a system design with improved communication reliability, utilizing Bluetooth mesh networks at the frontend and a fog-based cloud backend. The design is to handle dynamic data traffic load variations, such as those occurring in suddenly crowded hospital environments during epidemics, where excess patients may be placed in random locations, including corridors. With the widespread adoption of BLE in smartphones, patient endpoints can be easily set up in our system during emergencies. Testbed experiments and measurements confirm that frames from different BLE nodes in our model can reach edge devices (e.g., nurse stations) through three different operating modes. The CoAP-assisted Fog (CaF) mode offers delivery guarantees with high success rates than that of the regular BLE mesh mode.

Consensus algorithms are crucial for data consistency in distributed systems with member nodes spread across the networks. Traditionally, consensus algorithms, such as Paxos and Raft, may not scale well for large memberships on the Internet. In this paper, we propose a geographically wide-area large-scale consensual system that runs on two types of communication media: (1) the wireline connections on the Internet, and (2) wireless broadcast channel through a High Altitude Platform Station (HAPS) system. Each node in the system is a physical location on Earth’s surface. A data center inside a building is one such example. We call our system a Channel Heterogeneous Consensual System (CHCS), and it can be classified as a type of Vertical Heterogeneous Network (VHetNet). The Channel Heterogeneous Consensus Protocol (CHCP) runs on the CHCS to facilitate consensus agreements for data values across all member nodes. Through simulations, the consensus agreements are achievable through the CHCP, and the system scales well for a large number of member nodes. In general, there are situations such as potential data losses through wireless communication channels, and there are cases where the HAPS airships may fail to reach the target recipients on Earth. In our system, there is backup wireline connectivity among all nodes running the regular Raft that can cover those nodes with lost messages. Through our simulations, especially for cases with large membership numbers, it is shown that the times required by the CHCS to achieve consensus agreements are in general faster than those of the regular wireline Raft algorithm.

With meta-learning, models are trained on multiple tasks, and resulting trained models are expected to be capable of “ learning” new tasks effectively. MAML (Model Agnostic Meta-Learning) was one such early design which allowed models to reuse learned features, but with limited performance. Pre-training is another known method to improve the performance of a final trained model. Pre-train starts by assisting models to seek better initialization points, thus offering better feature representations. In this paper, we propose doublePT (double-pretrain) which is a two-stage pre-training method with goals to 1) reduce the number of parameters, 2) acquire better feature representations, and 3) achieve competitive overall performances in different benchmark measurements. For the first stage operation, we use a universal pre-training model to capture general features from a large dataset. In the second stage, we propose to use MAML to fine-tune the pre-training model to enhance the feature adaptability. Since the first-stage pre-training model has already learned general feature representations, it reduces the training activities for the second-stage fine-tuning operations, and enables better feature extractions in new tasks. Validated through our experiments, we find that our proposed doublePT approach fine-tunes across different tasks, and performs better than that of one-stage pre-training approach. Upon combining doublePT and DINOv2, and comparing to the latest PMF meta-learning design, the number of parameters required by the PMF pipelining model needs 304.8% more parameters than in our proposed DINOv2+doublePT model design. Performance-wise, the DINOv2+doublePT also has the best accuracies across different benchmark measurements.

Sensors in wireless sensor network (WSN) usually spatially spread across geographical locations. They may be placed randomly or in an initially organized manner to cooperatively monitor certain physical or environmental phenomena. They have limited transmission powers due to their small sizes and battery constraints. Some sensors may not be able to send data directly to the sink for processing and analysis. This has led to, for example, the design of tree-based structure for delivering data over multiple hops to reach the sink. In this paper, a novel lifetime extending heuristic (MLC-X) is proposed for tree-based multi-level clustered wireless sensor network. Duties of nodes at bottlenecks in tree are modified for sustaining longer network lifetime. And the simulation results indicate that the heuristic can successfully extend life spans of sensor networks.

IEEE 802.11 wireless local area networks (WLANs) have been considered as potential solutions for constructing infrastructures of wireless mesh networks (WMNs). For connections going through multiple wireless hops, the interference and transmission range issue at physical layer, terminal problems at link layer, and unsatisfactory performance at transport layer may render 802.11 Medium Access Control (MAC) protocol inappropriate for wireless mesh networks. In fact, the hidden and exposed terminal problems magnify if interference range is multiple times wider than transmission range. The multi-radio multi-channel WLAN is a possible solution for multi-hop mesh networks. In this paper, a multi-channel MAC protocol is extended to combat the interference range and the hidden terminal problems for multi-hop wireless mesh networks. Besides, thorough simulation results indicate that the design offers excellent throughput performance in WLAN environments.

Distributed hash tables (DHTs), based on consistent hashing, offer efficient lookup services for decentralized distributed systems. DHTs operate efficiently to handle large number of network nodes with continual node arrivals, departures, and failures. Upon addressing the crucial issues of communication efficiency and offering load balancing in dynamic networking environments, DHTs are the essential components for building structured peer-to-peer (P2P) overlay networks. Although structured overlays improve data availability and consistency, they do not provide strong semantics on distributed data mutual exclusion operations. For a robust network operating system, it is essential to provide atomic data access semantic services. In this paper, a distributed semaphore (DISEM) mechanism is proposed, and it is designed on top of a dynamic structured overlay. The proposed design circumvents the availability and consistency issues. Independent of any underlying overlay algorithms, DISEM provides a tunable level of data availability and consistency, while offering fault tolerance and reliable delivery services. A testbed prototype has been implemented to validate the mutual exclusiveness of networked replicas under different traffic loadings. The measured results indicate that DISEM offers high mutual exclusive access rates under different networking conditions.

Large-scale online applications such as Massively Multiplayer Online Games (MMOGs) require large amount of computing resources that support many players interacting simultaneously. Cluster computing is the technology mostly used by online game designing firms. Cluster computing is limited by the number and types of computers it can manage, but these computers are usually in the same geographical location. On the other hand, Grid computing offers large-scale high performance distributed computing which connects various types of computing resources on the Internet. In this paper, we design a Grid computing platform called the Massively Multi-user Online Platform (MMOP). The objectives of this proposed design are to offer scalability, flexibility, and simplicity to the development processes of distributed applications. MMOP allows executions of applications based on specified policy rules with dynamic addition of computing resources at run-time. Each application is managed separately, and multiple large-scale applications can share a single computing architecture. An online game has been built to test the functional behavior of the MMOP. From the simulation results, the MMOP has demonstrated as a high performance and scalable computing architecture.

Pervasive computing is the next generation computing environment with information and communication technology everywhere, for everyone, at any time. There are many methods proposed to reach the ultimate design goals of pervasive computing. Most of them are still at the hypothesis and early stages. In this paper, an initial investigation has been carried out to use active networks as the operating platforms for pervasive computing. Upon taking advantage of the active network paradigm, which offers flexibility and extensibility within networks, the active pervasive network infrastructure (APNI) framework is proposed. An additional feature known as ‘adaptiveness’ is recommended to work in pervasive computing. Primarily, the integrity of information content can be retained and adapted to available network resources, even if an end user moves to an unfavourable network environment as long as the user approves and accepts appropriate modifications of receiving content. Furthermore, information transfer can be sustained through different extensible mechanisms on the network layer in active networks. As a result, information can be sent through the APNI with low packet loss rate. The functional objectives of proposed designs in the paper have been verified through thorough experiments in an active network prototype.

Micro-electro-mechanical system (MEMS) is one of the few commercial platforms for building optical switches. 2D MEMS L-switching matrix has been introduced recently to double sizes of 2D MEMS crossbar switches. The sizes of the switches are mainly limited by the Gaussian signal loss associated path difference. Though the design of L-switching matrix improves system scalability but it suffers internal blocking problem. In this paper, a rearrangeably nonblocking algorithm will be presented. Moreover, two enhanced designs are proposed to improve the overall system performance of the L-switching matrix. They are the staircase switching mechanism and redundant switching system. With the improved internal blocking probability can be minimized. Consequently, the L-switching probability can be minimized. Consequently, the L-switching matrix performs similar to an optical switching fabric with wide-sense nonblocking property.

Explosion in Internet applications has stimulated active research activities in expanding the capability of the current telecommunication networks. These activities include implementing faster electronics to process the higher data bit rates, or developing wavelength division multiplexing (WDM) techniques, and novel optical networks components to increase the information carrying capacity of the optical networks. As the data bit rates increases, it will become increasingly difficult to implement an electronic switching fabric solution. It is known that the information carrying laser beams should be dealt with at the optical level. One of most promising optical network components is micromachined optical cross connect switches. We present a new crossbar switching design methodology that decreases the number of mirrors and electrodes needed while maintaining the same non-blocking port switching capability. More importantly, this new architecture also reduces the distance of free-space propagation of light beams, thus reducing the loss due to Guassian-beam divergence during free-space propagation of light.

The performance of multimedia applications on the Internet relates to network bandwidth availability, packet loss, and delay factors, as well as the human perceptions. Quality also links to the network delivery cost that a user is willing to pay. The paper presents a protocol, service bidding protocol, that allows multimedia applications to negotiate and re-negotiate on resource reservation with the networks. It enables a flexibility that applications can adapt to different network resources on-the-fly. This protocol works with the differentiated service (DiffServ) model, but can be extended easily to other service models. An implementation is constructed to evaluate the design performance.

The concept of programmable nodes and active networks introduces programmability into communication networks. Code and data can be sent and modified on their routes to the destinations. Various research groups have designed and implemented their own design platforms. Each design has its own benefits and drawbacks. Moreover, there exists an interoperability problem among platforms. As a result, we introduce a concept that is similar to network socket programming. We intentionally establish a set of simple interfaces for programming active applications. This set of interfaces, known as active network socket programming (ANSP), will be working on top of all other execution environments in the future. Therefore, the ANSP offers a concept that is similar to "write once, run everywhere". It is an open programming model that active applications can work on all execution environments. It solves the heterogeneity within active networks. This is especially useful when active applications need to access all regions within a heterogeneous network to deploy a special service at critical points or to monitor the performance of the entire networks. Instead of introducing a new platform, our approach provides a thin, transparent layer on top of existing environments that can be easily installed for all active applications.

This paper focuses on designing a large N/spl times/N high-performance broad-band ATM switch. Despite advances in architectural designs, practical switch dimensions continue to be severely limited by both the technological and physical constraints of packaging. Here, we focus on augmentation in a "single-switch" design: we provide ways to construct arbitrarily large switches out of modest-size components and retain overall delay/throughput performance. We propose a growable switch architecture based on several key principles: 1) the knockout principle exploits the statistical behavior of cell arrivals, and thereby reduces the interconnect complexity; 2) output queueing yields the best possible delay/throughput performance; 3) distributed control in routing (multicast) cells through the interconnect fabric without internal path conflicts; and 4) simple basic building blocks facilitate scalability. Other attractive features of the proposed architecture include: 1) intrinsic broadcast and multicast capabilities; 2) built-in priority sorting functionality; and 3) the guarantee of first-in, first-out cell sequence, To achieve 10/sup -14/ cell loss probability, only maximum size 32/spl times/16 basic building modules are required, and no crossover interconnects exist between modules in a three-dimensional configuration.

We present a novel design concept for constructing high-speed output port controllers for ATM switches, and statistical multiplexers. The newly proposed concept provides an external framework in which the internal hardware designs can be modified to achieve a specific quality of service. Two distributed control designs are provided, namely the fully shared buffer and partially shared buffer architectures. The fully shared buffer architecture can provide the push-out mechanism or complete buffer sharing queueing discipline which gives the best loss and delay performance for an incoming cell stream with static priority.