Existing studies on federated learning (FL) are mostly focused on system orchestration forstatic snapshotsof the network and makingstatic control decisions(e.g., spectrum allocation). However, real-world wireless networks are susceptible totemporal variationsof wireless channel capacity and users’ datasets. In this paper, we study the impacts of the dynamics: 1) wireless channels and 2) users’ datasets on the FL execution. The former is captured by introducing a set of discrete time events while the latter is characterized by a novelordinary differential equationand the metric ofdynamic model drift, formulated via apartial differential inequality, drawing concrete analytical connections between the dynamics of users’ datasets and FL accuracy. We then proposedynamiccooperative FLwith dedicatedMAC schedulers (DCLM), exploiting the unique features of open radio access network (O-RAN) to execute FL.DCLMentails: 1) a hierarchical device-to-device (D2D)-assisted model training; 2) dynamic control decisions through dedicated O-RAN MAC schedulers; and 3) asymmetric user selection. We provide extensive theoretical analysis to study the convergence ofDCLMand then aim to optimize its degrees of freedom (e.g., user selection and spectrum allocation) through a non-convex optimization problem. We develop a systematic and generic approach to obtain the solution for this problem. We finally show the efficiency ofDCLMvia numerical simulations and provide a series of future directions.

We study how to schedule data sources in a wireless time-sensitive information system with multiple heterogeneous and unreliable channels to minimize the total expected Age-of-Information (AoI). Although one could formulate this problem as a discrete-time Markov Decision Process (MDP), such an approach suffers from the curse of dimensionality and lack of insights. For single-channel systems, prior studies have developed lower-complexity solutions based on the Whittle index. However, Whittle index has not been studied for systems with multiple heterogeneous channels, mainly because indexability is not well defined when there are multiple dual cost values, one for each channel. To overcome this difficulty, we introduce new notions of partial indexability and partial index, which are defined with respect to one channel’s cost, given all other channels’ costs. We then combine the ideas of partial indices and max-weight matching to develop a Sum Weighted Index Matching (SWIM) policy, which iteratively updates the dual costs and partial indices. The proposed policy is shown to be asymptotically optimal in minimizing the total expected AoI, under a technical condition on a global attractor property. We also propose an interpolation-based algorithm to quickly compute (approximate) partial indices in real time. Extensive performance simulations demonstrate that the proposed policy offers significant gains over conventional approaches by achieving a near-optimal AoI. Further, the notion of partial index is of independent interest and could be useful for other problems with multiple heterogeneous resources.

In this paper, we conduct a measurement study on operational 5G networks deployed across different frequency bands (mmWave and sub-6GHz) and server locations (mobile edge and Internet cloud). Specifically, we assess 5G performance in both uplink and downlink across multiple operators’ networks. We then carry out extensive comparisons of transport-layer protocols using ten different algorithms in full-fledged 5G networks, including an edge computing environment. Finally, we evaluate representative mobile applications over the 5G network with and without edge servers. Our comprehensive measurements provide several insights that affect the experience of 5G users: (i) With a 5G edge server, existing TCP congestion control algorithms can achieve throughput up to 1.8Gbps with only a single flow. (ii) The maximum TCP receive buffer size, which is set by off-the-shelf 5G phones, can limit the throughput performance of 5G networks, which is not observed in 4G LTE-A networks. (iii) Despite significant latency gains in download-centric applications, the 5G edge service provides limited benefits to CPU-intensive tasks or those that use significant uplink bandwidth. To our knowledge, this is the first measurement-driven understanding of 5G edge computing “in the wild,” which can provide an answer to how edge computing would perform in real 5G networks.

Network sharing is widely accepted as a cost effective approach for mobile network deployment. It remains uncertain, however, how regulators will evaluate network sharing agreements (NSA) for future networks in the context of the current competition law. For example, 5G mobile network operators (MNOs) seeking to enter NSAs may risk legal challenges, as regulators have not given MNOs sufficient guidance for self-evaluation of their NSAs. One way for MNOs to reduce the risk of legal challenge is to avoid sharing variable costs in the NSA. However, constraining costs to be non-variable (i.e., fixed) rules out the use of most pricing mechanisms that have been widely adopted for dynamic resource trading between MNOs. In this article, we propose a network sharing framework to allow dynamic resource sharing without the use of resource pricing. To incentivize sharing without pricing, our framework presents sharing as a means for MNOs to differentiate services and better compete in the service market for profit. We evaluate our framework in a duopoly market model and demonstrate the economic and regulatory viability of our framework.

We propose cooperative edge-assisted dynamic federated learning (CE-FL).CE-FLintroduces a distributed machine learning (ML) architecture, where data collection is carried out at the end devices, while the model training is conducted cooperatively at the end devices and the edge servers, enabled via data offloading from the end devices to the edge servers through base stations.CE-FLalso introduces floating aggregation point, where the local models generated at the devices and the servers are aggregated at an edge server, which varies from one model training round to another to cope with the network evolution in terms of data distribution and users’ mobility.CE-FLconsiders the heterogeneity of network elements in terms of communication/computation models and the proximity to one another.CE-FLfurther presumes a dynamic environment with online variation of data at the network devices which causes a drift at the ML model performance. We model the processes taken duringCE-FL, and conduct analytical convergence analysis of its ML model training. We then formulate network-awareCE-FLwhich aims to adaptively optimize all the network elements via tuning their contribution to the learning process, which turns out to be a non-convex mixed integer problem. Motivated by the large scale of the system, we propose a distributed optimization solver to break down the computation of the solution across the network elements. We finally demonstrate the effectiveness of our framework with the data collected from a real-world testbed.

A new multiple-input multiple-output (MIMO) receiver scheme for practical binary codes is proposed that provides consistent gains over conventional linear receivers. We first develop a practical successive integer forcing (IF) scheme based on practical binary codes rather than lattice codes. We then present the successive cancellation integer forcing (SC-IF) scheme, which combines and enhances successive IF and minimum mean squared error successive interference cancellation (MMSE-SIC). In this scheme, the receiver first decides whether individual decoding or IF sum decoding is appropriate for each data stream, and then conducts successive IF sum decoding only for selected streams while decoding the remaining streams using MMSE-SIC. The proposed SC-IF methodology mitigates the performance loss caused by mismatched IF filtering in fading channels, while attenuating the noise amplification caused by MMSE filtering. Extensive link-level simulations demonstrate that the proposed successive IF significantly improves the basic IF, and the SC-IF improves both the successive IF and MMSE-SIC, offering uniform improvements over conventional linear receivers for most channel correlation and variation parameters and modulation orders at comparable computational costs. These results illustrate the viability of SC-IF as a fundamental building block for high-performance MIMO receivers in 5G-Advanced and/or subsequent-generation communication systems.

Edge computing promises to exploit underlying computation resources closer to users to help run latency-sensitive applications such as augmented reality and video analytics. However, one key missing piece has been how to incorporate personally owned, unmanaged devices into a usable edge computing system. The primary challenges arise due to the heterogeneity, lack of interference management, and unpredictable availability of such devices. In this paper we propose an orchestration framework IBDASH, which orchestrates application tasks on an edge system that comprises a mix of commercial and personal edge devices. IBDASH targets reducing both end-to-end latency of execution and probability of failure for applications that have dependency among tasks, captured by directed acyclic graphs (DAGs). IBDASH takes memory constraints of each edge device and network bandwidth into consideration. To assess the effectiveness of IBDASH, we run real application tasks on real edge devices with widely varying capabilities. We feed these measurements into a simulator that runs IBDASH at scale. Compared to three state-of-the-art edge orchestration schemes and two intuitive baselines, IBDASH reduces the end-to-end latency and probability of failure, by 14% and 41% on average respectively. The main takeaway from our work is that it is feasible to combine personal and commercial devices into a usable edge computing platform, one that delivers low and predictable latency and high availability.

Error correcting codes are a fundamental component in modern day communication systems, demanding extremely high throughput, ultra-reliability and low latency. Recent approaches using machine learning (ML) models as decoders offer both improved performance and great adaptability to unknown environments, where traditional decoders struggle. We introduce a general framework to further boost the performance and applicability of ML models. We propose to combine ML decoders with a competing discriminator network that tries to distinguish between codewords and noisy words, and, hence, guides the decoding models to recover transmitted codewords. Our framework is game-theoretic, motivated by generative adversarial networks (GANs), with the decoder and discriminator competing in a zero-sum game. The decoder learns to simultaneously decode and generate codewords while the discriminator learns to tell the difference between decoded outputs and codewords. Thus, the decoder is able to decode noisy received signals into codewords, increasing the probability of successful decoding. We show a strong connection of our framework with the optimal maximum likelihood decoder by proving that this decoder defines a Nash equilibrium point of our game. Hence, training to equilibrium has a good possibility of achieving the optimal maximum likelihood performance. Moreover, our framework does not require training labels, which are typically unavailable during communications, and, thus, seemingly can be trained online and adapt to channel dynamics. To demonstrate the performance of our framework, we combine it with recent neural decoders and show improved performance compared to the original models and traditional decoding algorithms on various codes.

Edge computing has revolutionized the world of mobile and wireless networks world thanks to its flexible, secure, and performing characteristics. Lately, we have witnessed the increasing use of it to make more performing the deployment of machine learning (ML) techniques such as federated learning (FL). FL was debuted to improve communication efficiency compared to conventional distributed machine learning (ML). The original FL assumes a central aggregation server to aggregate locally optimized parameters and might bring reliability and latency issues. In this paper, we conduct an in-depth study of strategies to replace this central server by a flying master that is dynamically selected based on the current participants and/or available resources at every FL round of optimization. Specifically, we compare different metrics to select this flying master and assess consensus algorithms to perform the selection. Our results demonstrate a significant reduction of runtime using our flying master FL framework compared to the original FL from measurements results conducted in our EdgeAI testbed and over real 5G networks using an operational edge testbed.

We study how to schedule data sources in a wireless time-sensitive information system with multiple heterogeneous and unreliable channels to minimize the total expected Age-of-Information (AoI). Although one could formulate this problem as a discrete-time Markov Decision Process (MDP), such an approach suffers from the curse of dimensionality and lack of insights. For single-channel systems, prior studies have developed lower-complexity solutions based on the Whittle index. However, Whittle index has not been studied for systems with multiple heterogeneous channels, mainly because indexability is not well defined when there are multiple dual cost values, one for each channel. To overcome this difficulty, we introduce new notions of partial indexability and partial index, which are defined with respect to one channel's cost, given all other channels' costs. We then combine the ideas of partial indices and max-weight matching to develop a Sum Weighted Index Matching (SWIM) policy, which iteratively updates the dual costs and partial indices. The proposed policy is shown to be asymptotically optimal in minimizing the total expected AoI, under a technical condition on a global attractor property. Extensive performance simulations demonstrate that the proposed policy offers significant gains over conventional approaches by achieving a near-optimal AoI. Further, the notion of partial index is of independent interest and could be useful for other problems with multiple heterogeneous resources.

Running intensive compute tasks across the fifth generation mobile network of edge devices introduces distributed computing challenges: edge devices are heterogeneous in the compute, storage, and communication capabilities; and can exhibit unpredictable straggler effects and failures. In this work, we propose an error-correcting-code inspired strategy to execute computing tasks in edge computing environments, which is designed to mitigate variability in response times and errors caused by edge devices' heterogeneity and lack of reliability. Unlike prior coding approaches, we incorporate partially unfinished coded tasks into our computation recovery, which allows us to achieve smooth performance degradation with low-complexity decoding when the coded tasks are run on edge devices with a fixed deadline. By further carrying out coding on edge devices as well as a master node, the proposed computing scheme also alleviates communication bottlenecks during data shuffling and is amenable to distributed implementation in a highly variable and limited network. Such distributed encoding forces us to solve new decoding challenges. Using a representative implementation based on federated multi-task learning frameworks, extensive performance simulations are carried out, which demonstrate that the proposed strategy offers significant gains in latency and accuracy over conventional coded computing schemes.

We study low-overhead uplink multi-access algorithms for massive Internet-of-Things (IoT) that can exploit the MIMO performance gain. Although MIMO improves system capacity, it usually requires high overhead due to Channel State Information (CSI) feedback, which is unsuitable for IoT. Recently, a Pseudo-Random Beam-Forming (PRBF) scheme was proposed to exploit the MIMO performance gain for uplink IoT access with uniform channel and load, without collecting CSI at the BS. For non-uniform channel and load, new adaptive beamselection and random-access algorithms are needed to efficiently utilize the system capacity with low overhead. Most existing algorithms for a related multi-channel scheduling problem require each node to at least know some information of the queue length of all contending nodes. In contrast, we propose a new Low-overhead Multi-Channel Joint Channel-Assignment and Random-Access (L-MC-JCARA) algorithm that reduces the overhead to be independent of the number of interfering nodes. A key novelty is to let the BS estimate the total backlog in each contention group by only observing the random-access events, so that no queue-length feedback is needed from IoT devices. We prove that L-MC-JCARA can achieve at least `0.24`` of the capacity region of the optimal centralized scheduler for the corresponding multi-channel system.

A pseudo-random beamforming (PRBF) based random access (RA) system is proposed to enable uplink (UL) machine-type communications (MTC) with ultra low signaling overheads. Specifically, a pseudo random (PR) sequence is used as public information to coordinate the beamforming vectors used at the base station (BS) and the devices. Within the coherence time window, each device distributively determines in advance the ''good'' time slots and receiving beams for transmission. This UL protocol reduces the overheads due to the feedback of channel state information and the control signals for centralized scheduling. This paper derives the throughput and user scaling of the proposed M- PRBF-CA protocol for achieving spatial multiplexing gain, under both an i.i.d. slow fading channel and a correlated slow fading channel. Our simulation results confirm the analysis in both fading channel models.

A new coding scheme for general N -node relay networks is presented for unicast, multicast, and broadcast. The proposed distributed decode-forward scheme combines and generalizes Marton coding for single-hop broadcast channels and the Cover-El Gamal partial decode-forward coding scheme for three-node relay channels. The key idea of the scheme is to precode all the codewords of the entire network at the source by multicoding over multiple blocks. This encoding step allows these codewords to carry partial information of the messages implicitly without complicated rate splitting and routing. This partial information is then recovered at the relay nodes and forwarded further. For N-node Gaussian unicast, multicast, and broadcast relay networks, the scheme achieves within 0.5N bits from the cutset bound, and thus from the capacity (region), regardless of the network topology, channel gains, or power constraints. Roughly speaking, distributed decode-forward is dual to noisy network coding, which generalized compress-forward to unicast, multicast, and multiple access relay networks.

The sliding-window superposition coding scheme aims to mitigate intercell interference at the physical layer by achieving the simultaneous decoding performance with point-to-point channel codes, low- complexity decoding, and minimal coordination overhead. The associated sliding-window coded modulation (SWCM) scheme can be readily implemented using standard off-the-shelf codes, such as the standard LTE turbo code, and tracks the information-theoretical performance guarantee of sliding-window superposition coding. This paper investigates how the basic SWCM scheme performs for the Ped-B fading interference channel model and proposes several improvements in transceiver design, such as soft decoding, input bit-mapping and layer optimization, and power control. Our enhanced SWCM scheme achieves the rates higher than those of the basic SWCM scheme by 10% to 20%, which already shows a significant gain over existing schemes that ignore modulation or coding information of interfering signals. This result confirms the potential of SWCM as a basic building block for physical-layer interference management in 5G and subsequent generations of cellular networks.

A new coding scheme for multicasting a message over a general relay network is presented that extends both network coding for graphical networks by Ahlswede, Cai, Li, and Yeung, and partial decode-forward for relay channels by Cover and El Gamal. For the N-node Gaussian multicast network, the scheme achieves within 0.5N bits from the capacity, improving upon the best known capacity gap results. The key idea is to use multicoding at the source as in Marton coding for broadcast channels. Instead of recovering a specific part of the message as in the original partial decode-forward scheme, a relay in the proposed distributed decode-forward scheme recovers an auxiliary index that implicitly carries some information about the message and forwards it in block Markov coding. This scheme can be adapted to broadcasting multiple messages over a general relay network, extending and refining a recent result by Kannan, Raja, and Viswanath.

A new coding scheme for broadcasting multiple messages over a general relay network is presented. The proposed distributed decode-forward scheme combines Marton coding for single-hop broadcast channels and partial decode-forward for relay channels by Cover and El Gamal. For the N-node Gaussian broadcast relay network, the scheme achieves within 0.5N bits from the capacity region, extending and refining a recent result by Kannan, Raja, and Viswanath. The main idea of the scheme is to precode all the codewords initially at the source and to decode and forward parts of them on the fly at the relays.

In critical distortion range, the MSE rate-distortion function of time-discrete stationary Gaussian first-order autoregressive source is equal to that of related time-discrete i.i.d. Gaussian source. For 0cit is necessary to provide additional information of non-negligible positive rate in order to obtain a D-admissible code for the original source via the R-D coding of the innovations process and additional post-processing at the decoder. In this scenario, we provide an explicit expression of additional description rate about the original source to find the degree of suboptimality of sending a lossy version of the innovations process in Gauss-Markov rate-distortion. It is shown that additional description rate is monotone increasing on alpha isin [0,1) and is constant on all critical distortion range

In the critical range Oc, the MSE rate-distortion function of a time-discrete stationary autoregressive Gaussian source is equal to that of a related time-discrete i.i.d. Gaussian source. This suggests that perhaps an optimum encoder should compute the related memoryless sequence from the given source sequence with memory and then use a code of rate R(D) to convey the memoryless sequence to the decoder with an MSE of D. In this scenario, the question is, "for D les Dccan a D-admissible code for the original source be obtained via the R-D coding of the innovations process and additional post-processing at the decoder without having to provide any additional information of positive rate?" We show that the answer of this question often is "No"