ICC Lab

The sixth-generation (6G) wireless network landscape is evolving toward enhanced programmability, virtualization, and intelligence to support heterogeneous use cases. The O-RAN Alliance is pivotal in this transition, introducing a disaggregated architecture and open interfaces within the 6G network. Our paper explores an intelligent traffic steering (TS) scheme within the Open radio access network (RAN) architecture, aimed at improving overall system performance. Our novel TS algorithm efficiently manages diverse services, improving shared infrastructure performance amid unpredictable demand fluctuations. To address challenges like varying channel conditions, dynamic traffic demands, we propose a multi-layer optimization framework tailored to different timescales. Techniques such as long-short-term memory (LSTM), heuristics, and multi-agent deep reinforcement learning (MADRL) are employed within the non-real-time (non-RT) RAN intelligent controller (RIC). These techniques collaborate to make decisions on a larger timescale, defining custom control applications such as the intelligent TS-xAPP deployed at the near-real-time (near-RT) RIC. Meanwhile, optimization on a smaller timescale occurs at the RAN layer after receiving inferences/policies from RICs to address dynamic environments. The simulation results confirm the system’s effectiveness in intelligently steering traffic through a slice-aware scheme, improving eMBB throughput by an average of 99.42% over slice isolation.

Federated Learning (FL) plays a pivotal role in enabling artificial intelligence (AI)-based mobile applications in mobile edge computing (MEC). However, due to the resource heterogeneity among participating mobile users (MUs), delayed updates from slow MUs may deteriorate the learning speed of the MEC-based FL system, commonly referred to as the straggling problem. To tackle the problem, this work proposes a novel privacy-preserving FL framework that utilizes homomorphic encryption (HE) based solutions to enable MUs, particularly resource-constrained MUs, to securely offload part of their training tasks to the cloud server (CS) and mobile edge nodes (MENs). Our framework first develops an efficient method for packing batches of training data into HE ciphertexts to reduce the complexity of HE-encrypted training at the MENs/CS. On that basis, the mobile service provider (MSP) can incentivize straggling MUs to encrypt part of their local datasets that are uploaded to certain MENs or the CS for caching and remote training. However, caching a large amount of encrypted data at the MENs and CS for FL may not only overburden those nodes but also incur a prohibitive cost of remote training, which ultimately reduces the MSP’s overall profit. To optimize the portion of MUs’ data to be encrypted, cached, and trained at the MENs/CS, we formulate an MSP’s profit maximization problem, considering all MUs’ and MENs’ resource capabilities and data handling costs (including encryption, caching, and training) as well as the MSP’s incentive budget. We then show that the problem is convex and can be efficiently solved using an interior point method. Extensive simulations on a real-world human activity recognition dataset show that our proposed framework can achieve much higher model accuracy (improving up to 24.29%) and faster convergence rate (by 2.86 times) than those of the conventional FedAvg approach when the straggling probability varies between 20% and 80%. Moreover, the proposed framework can improve the MSP’s profit up to 2.84 times compared with other baseline FL approaches without MEN-assisted training.

We consider unmanned aerial vehicle (UAV)-enabled wireless systems where downlink communications between a multi-antenna UAV and multiple users are assisted by a hybrid active-passive reconfigurable intelligent surface (RIS). We aim at a fairness design of two typical UAV-enabled networks, namely the static-UAV network where the UAV is deployed at a fixed location to serve all users at the same time, and the mobile-UAV network which employs the time division multiple access protocol. In both networks, our goal is to maximize the minimum rate among users through jointly optimizing the UAV’s location/trajectory, transmit beamformer, and RIS coefficients. The resulting problems are highly nonconvex due to a strong coupling between the involved variables. We develop efficient algorithms based on block coordinate ascend and successive convex approximation to effectively solve these problems in an iterative manner. In particular, in the optimization of the mobile-UAV network, closed-form solutions to the transmit beamformer and RIS passive coefficients are derived. Numerical results show that a hybrid RIS equipped with only 4 active elements and a power budget of 0 dBm offers an improvement of 38%–63% in minimum rate, while that achieved by a passive RIS is only about 15%, with the same total number of elements.

In this paper, we address a dynamic network resource selection problem for mobile users in a rate-splitting multiple access (RSMA)-enabled network by leveraging evolutionary games. Particularly, mobile users are able to locally and dynamically make their selection on orthogonal resource blocks (RBs), which are also considered as network resources (NRs), over time to achieve their desired utilities. Then, RSMA is used for each group of users selecting the same NR. With the use of RSMA, the main goal is to optimize the beamformers of the common and private messages for users in the same group to maximize their sum rate. The resulting problem is generally non-convex, and thus we develop a successive convex approximation (SCA)-based algorithm to efficiently solve it in an iterative fashion. To model the NR adaptation of users, we propose to use two evolutionary games, i.e . a traditional evolutionary game (TEG) and fractional evolutionary game (FEG). The FEG approach enables users to incorporate memory effects ( i.e . their past experiences) for their decision-making, which is more realistic than the TEG approach. We then theoretically verify the existence of the equilibrium of the proposed game approaches. Simulation results are provided to validate their consistency with the theoretical analysis and merits of the proposed approaches. They also reveal that, compared with TEG, FEG enables users to leverage past information for their decision-making, resulting in less communication overhead, while still guaranteeing convergence.

Open radio access network (ORAN) Alliance offers a disaggregated RAN functionality built using open interface specifications between blocks. To efficiently support various competing services, namely enhanced mobile broadband (eMBB) and ultra-reliable and low-latency (uRLLC), the ORAN Alliance has introduced a standard approach toward more virtualized, open, and intelligent networks. To realize the benefits of ORAN in optimizing resource utilization, this paper studies an intelligent traffic steering (TS) scheme within the proposed disaggregated ORAN architecture. For this purpose, we propose a joint intelligent traffic prediction, flow-split distribution, dynamic user association, and radio resource management (JIFDR) framework in the presence of unknown dynamic traffic demands. To adapt to dynamic environments on different time scales, we decompose the formulated optimization problem into two long-term and short-term subproblems, where the optimality of the latter is strongly dependent on the optimal dynamic traffic demand. We then apply a long-short-term memory (LSTM) model to effectively solve the long-term subproblem, aiming to predict dynamic traffic demands, RAN slicing, and flow-split decisions. The resulting non-convex short-term subproblem is converted to a more computationally tractable form by exploiting successive convex approximations. Finally, simulation results are provided to demonstrate the effectiveness of the proposed algorithms compared to several well-known benchmark schemes.

Federated learning (FL) is capable of performing large distributed machine learning tasks across multiple edge users by periodically aggregating trained local parameters. To address key challenges of enabling FL over a wireless fog-cloud system (e.g., non-i.i.d. data, users’ heterogeneity), we first propose an efficient FL algorithm based on Federated Averaging (called FedFog ) to perform the local aggregation of gradient parameters at fog servers and global training update at the cloud. Next, we employ FedFog in wireless fog-cloud systems by investigating a novel network-aware FL optimization problem that strikes the balance between the global loss and completion time. An iterative algorithm is then developed to obtain a precise measurement of the system performance, which helps design an efficient stopping criteria to output an appropriate number of global rounds. To mitigate the straggler effect, we propose a flexible user aggregation strategy that trains fast users first to obtain a certain level of accuracy before allowing slow users to join the global training updates. Extensive numerical results using several real-world FL tasks are provided to verify the theoretical convergence of FedFog . We also show that the proposed co-design of FL and communication is essential to substantially improve resource utilization while achieving comparable accuracy of the learning model.

In-band full-duplex (FD) operation is practically more suited for short-range communications such as WiFi and small-cell networks, due to its current practical limitations on the self-interference cancellation. In addition, cell-free massive multiple-input multiple-output (CF-mMIMO) is a new and scalable version of MIMO networks, which is designed to bring service antennas closer to end user equipments (UEs). To achieve higher spectral and energy efficiencies (SE-EE) of a wireless network, it is of practical interest to incorporate FD capability into CF-mMIMO systems to utilize their combined benefits. We formulate a novel and comprehensive optimization problem for the maximization of SE and EE in which power control, access point-UE (AP-UE) association and AP selection are jointly optimized under a realistic power consumption model, resulting in a difficult class of mixed-integer nonconvex programming. To tackle the binary nature of the formulated problem, we propose an efficient approach by exploiting a strong coupling between binary and continuous variables, leading to a more tractable problem. In this regard, two low-complexity transmission designs based on zero-forcing (ZF) are proposed. Combining tools from inner approximation framework and Dinkelbach method, we develop simple iterative algorithms with polynomial computational complexity in each iteration and strong theoretical performance guaranteed. Furthermore, towards a robust design for FD CF-mMIMO, a novel heap-based pilot assignment algorithm is proposed to mitigate effects of pilot contamination. Numerical results show that our proposed designs with realistic parameters significantly outperform the well-known approaches (i.e., small-cell and collocated mMIMO) in terms of the SE and EE. Notably, the proposed ZF designs require much less execution time than the simple maximum ratio transmission/combining.

We consider a full-duplex (FD) multiuser system where an FD base station (BS) is designed to simultaneously serve both downlink (DL) and uplink (UL) users in the presence of half-duplex eavesdroppers (Eves). The problem is to maximize the minimum (max-min) secrecy rate (SR) among all legitimate users, where the information signals at the FD-BS are accompanied with artificial noise to debilitate the Eves’ channels. To enhance the max-min SR, a major part of the power budget should be allocated to serve the users with poor channel qualities, such as those far from the FD-BS, undermining the SR for other users, and thus compromising the SR per-user. In addition, the main obstacle in designing an FD system is due to the self-interference (SI) and co-channel interference (CCI) among users. We therefore propose an alternative solution, where the FD-BS uses a fraction of the time block to serve near DL users and far UL users, and the remaining fractional time to serve other users. The proposed scheme mitigates the harmful effects of SI, CCI, and multiuser interference, and provides system robustness. The SR optimization problem has a highly nonconcave and nonsmooth objective, subject to nonconvex constraints. For the case of perfect channel state information (CSI), we develop a low-complexity path-following algorithm, which involves only a simple convex program of moderate dimension at each iteration. We show that our path-following algorithm guarantees convergence at least to a local optimum. Then, we extend the path-following algorithm to the cases of partially known Eves’ CSI, where only statistics of CSI for the Eves are known, and worst-case scenario in which Eves can employ a more advanced linear decoder. The merit of our proposed approach is further demonstrated by extensive numerical results.

WiFi-based human pose estimation (HPE) has emerged as a promising alternative to conventional vision-based techniques, yet faces the high computational cost hindering its widespread adoption. This pa- per introduces a novel HPE-Li approach that harnesses multi-modal sen- sors (e.g. camera and WiFi) to generate accurate 3D skeletal in HPE. We then develop an efficient deep neural network to process raw WiFi signals. Our model incorporates a distinctive multi-branch convolutional neural network (CNN) empowered by a selective kernel attention (SKA) mechanism. Unlike standard CNNs with fixed receptive fields, the SKA mechanism is capable of dynamically adjusting kernel sizes according to input data characteristics, enhancing adaptability without increasing complexity. Extensive experiments conducted on two MM-Fi and WiPose datasets underscore the superiority of our method over state-of-the-art approaches, while ensuring minimal computational overhead, rendering it highly suitable for large-scale scenarios