Mengmeng Ren

This paper investigates a cell-free massive multiple-input multiple-output enabled multi-access edge computing (termed CF-MEC) system, where multiple users are served by multiple central processing units (CPUs) and their connected access points (APs), both of which are equipped with computation resources. For this system, a dynamic user-centric task offloading scheme is designed to provide seamless and efficient computation services for users. Based on this scheme, the joint optimization of user-centric AP clustering, edge server selection, communication and computation resources is formulated as a long-term problem to minimize the average energy consumption. The formulated problem is complicated non-convex due to the highly coupled time-varying discrete and continuous variables, resulting in high complexity and non-real-time to obtain the optimal solution. To tackle this challenging problem, we propose a multi-layer hierarchical multi-agent deep reinforcement learning (ML-HMADRL) based resource allocation algorithm. Specifically, the proposed algorithm incorporates a hierarchical structure with high, middle, and low-level agents that iteratively train the actor-critic networks to obtain discrete and continuous variables of the formulated problem. To further enhance the training effectiveness by leveraging the CF-MEC system, we design distinct actor-critic networks for the agents at different levels to facilitate centralized training and distributed execution. Simulation results validate the training stability of the proposed algorithm at each level, and demonstrate the superiority of the proposed algorithm over benchmark schemes in terms of the average energy consumption, providing a stable distributed framework for practical implementation in dynamic environments.

This paper develops an edge-device collaborative Generative Semantic Communications (Gen SemCom) framework leveraging pre-trained Multi-modal/Vision Language Models (M/VLMs) for ultra-low-rate semantic communication via textual prompts. The proposed framework optimizes the use of M/VLMs on the wireless edge/device to generate high-fidelity textual prompts through visual captioning/question answering, which are then transmitted over a wireless channel for SemCom. Specifically, we develop a multi-user Gen SemCom framework using pre-trained M/VLMs, and formulate a joint optimization problem of prompt generation offloading, communication and computation resource allocation to minimize the latency and maximize the resulting semantic quality. Due to the non-convex nature of the problem with highly coupled discrete and continuous variables, we decompose it as a two-level problem and propose a low-complexity swap/leaving/joining (SLJ)-based matching algorithm. Simulation results demonstrate significant performance improvements over the conventional semantic-unaware/ non-collaborative generation offloading benchmarks.

The high-dynamic characteristics of vehicular networks make it difficult to satisfy strict deterministic quality-of-service (QoS) conditions. Statistical QoS provisioning instead of deterministic QoS provisioning can provide an efficient way to satisfy delay-bounded QoS requirements of vehicular services. Different from the existing work, this paper considers the heterogeneity of statistical QoS provisioning for delay-sensitive services in Internet-of-Vehicles (IoV), where the overall delay violation probability is limited in each time slot for mission-critical applications, and the queue length bound along with its higher-order statistics is limited from a long-term perspective for infotainment services. A long-term resource optimization problem is formulated to minimize the average energy consumption with heterogeneous statistical latency guarantees. To provide a stable and fast solution, Lyapunov optimization method is first leveraged to transform the formulated stochastic problem to a series of short-term deterministic optimization sub-problems. Afterwards, a novel two-level hierarchical deep reinforcement learning (H-DRL) framework is presented, where the knowledge sharing is further introduced to realize the fast model training and resource decision-making with QoS performance guarantees. Simulation results verify the convergence speed and the training accuracy of the presented H-DRL framework, and it also shows the superiority of the proposal in the perspective of the end-to-end delay and the queue stability for heterogeneous delay-bounded services under high-dynamic environments of IoV.

This paper designs a helper-assisted offloading strategy in non-orthogonal multiple access enabled mobile edge computing systems, in order to guarantee the quality of service of the energy/delay-sensitive user equipments (UEs). To achieve a tradeoff between the energy consumption and the delay, we introduce a performance metric called energy-delay tradeoff . Aiming at the maximal energy-delay tradeoff minimization, the joint optimization of user association, resource block (RB) assignment, power allocation, task assignment, and computation resource allocation is formulated as a non-convex problem with coupled continuous and 0-1 variables. To tackle this challenging problem, we decompose it as a two-level problem. For the inner-level problem, an iterative parametric convex approximation (IPCA) algorithm is proposed. Then, based on the solution obtained from the inner-level problem, we model the outer-level problem as a four-sided matching problem, and then propose a low-complexity four-sided UE-RB-helper-server matching (FS-URHSM) algorithm. Theoretical analysis demonstrates that the IPCA algorithm can converge to a stationary KKT point and the FS-URHSM algorithm is guaranteed to converge to a stable matching with polynomial complexity. Simulation results demonstrate the superior performance of proposed algorithms in terms of the energy consumption and the delay.

Efficient scheduling of vehicle resources is of great significance to guarantee vehicle safety and to achieve a higher level of automated driving. Considering the performance fluctuations in data transmission and processing during driving, this paper proposes an adaptive data transmission and computation optimization scheme, where the concept of the Internet-of-Intelligence is introduced to improve the resource decision-making efficiency through knowledge sharing instead of data sharing among vehicles. Specifically, the joint optimization problem is formulated to minimize the long-term energy consumption with the consideration of the average queuing latency guarantees. To provide a stable and fast solution, Lyapunov optimization method is first leveraged to transform the formulated stochastic problem into a series of short-term deterministic optimization subproblems. Afterwards, both the optimization-based solution and the learning-based solution are presented to fully illustrate the performance advantages of Internet-of-Intelligence applied to vehicle networks. The former can output the global optimal solution by iteration, while the latter aims at accelerating the distributed optimization decision-making through building a fast deep reinforcement learning framework based on shared knowledge among vehicles. Simulation results show the advantages of the proposed scheme in stability, energy consumption, and latency, and it also verifies the convergence speed and training accuracy of the proposed fast deep reinforcement learning framework.

This paper designs a helper-assisted resource allocation strategy in non-orthogonal multiple access (NOMA)-enabled mobile edge computing (MEC) systems, in order to guarantee the quality of service (QoS) of the energy/delay-sensitive user equipments (UEs). To achieve a tradeoff between the energy consumption and the delay, we introduce a novel performance metric, called \emph{energy-delay tradeoff}, which is defined as the weighted sum of energy consumption and delay. The joint optimization of user association, resource block (RB) assignment, power allocation, task assignment, and computation resource allocation is formulated as a mixed-integer nonlinear programming problem with the aim of minimizing the maximal energy-delay tradeoff. Due to the non-convexity of the formulated problem with coupled and 0-1 variables, this problem cannot be directly solved with polynomial complexity. To tackle this challenge, we first decouple the formulated problem into a power allocation, task assignment and computation resource allocation (PATACRA) subproblem. Then, with the solution obtained from the PATACRA subproblem, we equivalently reformulate the original problem as a discrete user association and RB assignment (DUARA) problem. For the PATACRA subproblem, an iterative parametric convex approximation (IPCA) algorithm is proposed. Then, based on the solution obtained from the PATACRA subproblem, we first model the DUARA problem as a four-sided matching problem, and then propose a low-complexity four-sided UE-RB-helper-server matching (FS-URHSM) algorithm. Theoretical analysis demonstrates that the proposed algorithms are guaranteed to converge to stable solutions with polynomial complexity. Finally, simulation results are provided to show the superior performance of our proposed algorithm in terms of the energy consumption and the delay.

This letter investigates the cooperative computing in a non-orthogonal multiple access (NOMA)-enabled mobile edge computing (MEC) system consisting of one edge server, one user equipment (UE) and $M$ full-duplex helpers. For this system, we design an adaptive cooperative computing strategy that operates in one of two modes: the cooperative NOMA-MEC mode and the non-cooperative orthogonal multiple access (OMA)-MEC mode. For this strategy, we propose a novel two-stage full-duplex helper scheduling scheme to improve the successful completion probability (SCP). Theoretical analysis is carried out to derive a tightly approximate SCP and further demonstrate that the UE achieves a full diversity in high main-to-loop-self-interference ratio regime. Simulation results verify the accuracy of the theoretical analysis and demonstrate the advantages of the proposed strategy.

This paper investigates the application of cooperative non-orthogonal multiple access (NOMA) to a two-tier wireless caching network, where the cache servers and the users are spatially randomly located and the users request cacheable contents that are pushed in advance from a base station to the cache servers. By integrating cooperative NOMA into wireless caching, a popularity-oriented cooperative NOMA pushing (POC-NOMA-P) strategy is proposed for content pushing, while a dynamic cooperative NOMA delivery (DC-NOMA-D) strategy is proposed for content delivery. A new analytical framework is designed to comprehensively evaluate the performance of the proposed strategies. Particularly, unlike existing works that purely focused on the analysis of cache hit probability (CHP) and outage probability, our analytical framework further introduces content diversity gain and delivery diversity gain to characterize the performance gain achieved by the use of cooperative NOMA. Moreover, the tradeoff analysis between performance and energy consumption is also included in our analytical framework. To ensure the analysis accuracy and mathematical tractability, a new tight approximation is developed for the cumulative distribution function of channel gain between two spatially random nodes, with which theoretic results are approximately derived for various performance metrics of our analytical framework. Our analytical results demonstrate that compared with the non-cooperative counterparts, the proposed strategies are guaranteed to achieve higher CHP/lower delivery outage probability and largely reduce the energy consumption incurred by content pushing/delivery. Simulations verify the accuracy of our derived approximations for various performance metrics and demonstrate the superiority of POC-NOMA-P and DC-NOMA-D over existing strategies.