Hao Yin

This letter studies the latency–power trade-off of various multi-link operation (MLO) traffic-allocation mechanisms – probability-based, greedy, and single link (SL) – using ns-3 simulations under both symmetric and asymmetric link conditions. We jointly consider mean and 99th-percentile (p99) end-to-end (E2E) latency as well as average power consumption of multi-link devices (MLDs). Results show that the greedy scheme achieves the lowest latency in symmetric traffic scenarios by exploiting simultaneous multi-link access, albeit with moderately higher power consumption. Conversely, under asymmetric link quality or background load, the probability-based method effectively collapses to single-link operation, attaining both minimal latency and lowest power. Our analysis sheds important light on adaptive traffic allocation for latency-sensitive and power-constrained Wi-Fi 7 applications.

As a groundbreaking feature in IEEE 802.11be, multi-link operation (MLO) is expected to support emerging applications that are strongly delay-sensitive. A key to the effective use of MLO for such cases rests on the optimal allocation of application traffic across multiple links. Our initial simulation experiments in ns-3 reveal that the proposed traffic allocation policies in prior art are significantly sub-optimal in terms of achievable delay performance of multi-link devices (MLDs), particularly in the presence of legacy single-link devices (SLDs). In this work, we first develop a new analytical model for the mean end-to-end (E2E) delay, delay jitter and worst-case percentile latency performance of MLD-SLD-coexisting Wi-Fi 7 networks (largely unexplored to date) with saturated and unsaturated SLD traffic. Subsequently, the optimal traffic allocation strategies for minimizing the mean E2E delay and delay jitter are obtained and validated by ns-3 simulation results. It is shown that with the optimal policy, MLDs can achieve significantly better mean E2E delay, delay jitter, worst-case latency and delay cumulative distribution function (CDF) compared to existing solutions.

With the proliferation of edge computing, efficient AI inference on edge devices has become essential for intelligent applications such as autonomous vehicles and VR/AR. In this context, we address the problem of efficient remote object recognition by optimizing feature transmission between mobile devices and edge servers. We propose an optimization framework to tackle the challenges posed by dynamic channel conditions and device mobility in end-to-end communication systems. Our approach builds upon existing methods by leveraging a semantic knowledge base to drive multi-level feature transmission, accounting for temporal factors, state transitions, and dynamic elements throughout the transmission process. Additionally, we enhance the multi-level feature transmission policy by introducing an additional 5th-level edge-assisted semantic communication, which maximizes recognition performance by leveraging a large SKB on the edge server. Formulated as an online optimization problem, our framework aims to simultaneously minimize semantic loss and adhere to specified transmission latency thresholds. To achieve this, we design a soft actor-critic-based deep reinforcement learning system with a carefully designed reward structure for real-time decision-making. This approach overcomes the optimization difficulty of the NP-hard problem while fulfilling the optimization objectives. Numerical results showcase the superiority of our approach compared to traditional greedy methods across various system setups using open-source datasets.

Sidelink is a 3GPP-standardized technology that enables direct UE-to-UE communication. Initially designed for licensed 4G spectrum (Release 15), it expanded to 5G NR (Release 16, 17). However, due to the recent availability of large swaths of unlicensed spectrum (notably 6 GHz in U.S.), 3GPP has initiated studies to support sidelink operation in such unlicensed bands (denoted as SL-U). SL-U specifications ride on adaptation of 5G NR operation from licensed to unlicensed bands (NR-U). This paper provides a first analytical framework to quantify SL-U network performance in saturation conditions, for comparative evaluation to NR-U. A Markov chain model for NR-U is first presented, as a basis for necessary (significant) modifications needed to capture SL-U-specific features - the representation of gap states in the SL-U model to account for the restriction that SL-U transmissions only occur at SL slot boundaries. The resulting analysis highlights the differences and consequent (negative) impact of current 3GPP-defined SL-U channel access rulemaking; the analytical estimates are validated by results from a custom simulation of channel access. The primary takeaway is that the SL-U throughput under current 3GPP access rules suffers significantly relative to NR-U, and thus needs amelioration for successful marketplace adoption in the future.

Wi-Fi is evolving towards enhanced support for real-time applications such as video conferencing and virtual reality (VR) gaming. The latest IEEE 802.11be draft standard introduces various advancements including multi-link operation (MLO) at the medium access control (MAC) layer. Such multi-link operation (effectively carrier aggregation) is anticipated to increase data throughput, and reduce access delay. Notably, the recent release of the ns-3 network simulator (ns-3.41) incorporates an implementation of MLO, enabling novel simulation-based analysis on the potential impact of link scheduling strategies on network delay. This paper begins with validation of end-to-end delay for the MLO implementation in ns-3 in a baseline scenario, and then proceeds to explore the impact of different link allocation strategies. Our observations reveal that across diverse scenarios, the outcomes derived from ns-3 align consistently with our theoretical model.

IEEE 802.11be (Wi-Fi 7) introduces a new concept called multilink operation (MLO), which allows multiple Wi-Fi interfaces in different bands (2.4, 5, and 6 GHz) to work together to increase network throughput, reduce latency, and improve spectrum reuse efficiency in dense overlapping networks. To make the most of MLO, this article proposes a new data-driven resource allocation algorithm for the 11be network with the aid of an access point (AP) controller. To maximize network throughput, a network topology optimization problem is formulated for 11be network, which is solved by exploiting the totally unimodular property of the bipartite graph formed by the connection between AP and station (STA) in Wi-Fi networks. Subsequently, a proportional fairness algorithm is applied for radio link allocation, network throughput optimization considering the channel condition, and the fairness of the multilink device (MLD) data rate. The performance of the proposed algorithm on two main MLO implementations—multilink-multiradio (MLMR) with simultaneous transmission and reception (STR), and the interplay between multiple nodes employing them are evaluated through cross-layer (PHY-MAC) data rate simulation with PHY abstraction.

With the proliferation of edge computing, efficient AI inference on edge devices has become essential for intelligent applications such as autonomous vehicles and VR/AR. In this context, we address the problem of efficient remote object recognition by optimizing feature transmission between mobile devices and edge servers. We propose an online optimization framework to address the challenge of dynamic channel conditions and device mobility in an end-to-end communication system. Our approach builds upon existing methods by leveraging a semantic knowledge base to drive multi-level feature transmission, accounting for temporal factors and dynamic elements throughout the transmission process. To solve the online optimization problem, we design a novel soft actor-critic-based deep reinforcement learning system with a carefully designed reward function for real-time decision-making, overcoming the optimization difficulty of the NP-hard problem and achieving the minimization of semantic loss while respecting latency constraints. Numerical results showcase the superiority of our approach compared to traditional greedy methods under various system setups.

New features brought by 5 G New Radio V2X (NR-V2X) support multiple vehicular communication types (unicast, groupcast, and broadcast) to coexist in road scenarios. However, the current standard does not specify the resource scheduling approach for groupcast to support such a new feature, which may severely degrade its packet delivery performance impacted by other communication types. In this paper, we investigate the scheduling and resource allocation approaches for a groupcast-based application, vehicle platooning, under the environment characterizing this new feature. We first analyze two baseline resource allocation approaches, i.e., Semi-persistent Scheduling (SPS) stated in the 3 rd generation partnership project (3GPP) and Random Selection (RS), based on the metric of packet collision probability. Subsequently, considering the issues from baseline approaches, we develop an Improved Random Selection (IRS) scheme to decrease the collision probability. We further propose to employ Deep Deterministic Policy Gradient (DDPG) algorithm to overcome the impact of inter-vehicle collaboration in the platoon based on local information. A Monte Carlo simulator is then used to verify the analytical models' results. The numerical results show that IRS significantly mitigates the packet collision probability compared with the baselines. Meanwhile, DDPG outperforms IRS in terms of the packet collision probability as well as average scheduled delay and is more robust to the change in the environment.