Artificial Intelligence Generated Content (AIGC) powered by Generative Diffusion Models (GDMs) has emerged as a transformative paradigm for automated content creation. To satisfy the stringent latency requirements of AIGC services in many edge intelligence scenarios (e.g., smart cities), Mobile Edge Computing (MEC) provides critical computational support by deploying GDMs at edge servers (ES) close to end users. This paper investigates an MEC-enabled AIGC network comprising multiple ES, wireless access points (APs), and mobile users (UEs) with heterogeneous latency and accuracy demands. We formulate aJoint Communication Association and Computation Offloading (JCACO)game, where each UE strategically selects its serving AP, ES, and inference steps to minimize the overall service completion time while meeting accuracy constraints. The problem is challenging due to the network dynamics and the incomplete information. We prove that the JCACO game is apotential gameunder both complete and stochastic information settings, ensuring the existence of Nash Equilibrium (NE) in both cases. To derive the NE efficiently, we develop a distributedMulti-Agent Stochastic Learning(MASL) algorithm that provably converges to the NE with strict performance guarantees. Unlike conventional best-response schemes, MASL requires neither the knowledge of other players’ strategies nor global network information, making it fully distributed and adaptive to dynamic environments. We further provide a strict theoretical convergence analysis for MASL by usingOrdinary Differential Equations(ODEs). Simulation results demonstrate that MASL significantly reduces service completion time compared with benchmark methods while satisfying accuracy constraints, confirming its effectiveness and practicality for real-world MEC-enabled AIGC networks.

Mobile Edge Computing (MEC) is a promising approach for enhancing the quality-of-service (QoS) of AI-enabled applications in the B5G/6G era, by bringing computation capability closer to end-users at the network edge. In this work, we investigate the joint optimization of edge server (ES) deployment, service placement, and computation task offloading under the stochastic information scenario. Traditional approaches often treat these decisions as equal, disregarding the differences in information realization. However, in practice, the ES deployment decision must be made in advance and remain unchanged, prior to the complete realization of information, whereas the decisions regarding service placement and computation task offloading can be made and adjusted in real-time after information is fully realized. To address such temporal coupling between decisions and information realization, we introduce the stochastic programming (SP) framework, which involves a strategic-layer for deciding ES deployment based on (incomplete) stochastic information and a tactical-layer for deciding service placement and task offloading based on complete information realization. The problem is challenging due to the different timescales of two layers' decisions. To overcome this challenge, we propose a multi-timescale SP framework, which includes a large timescale (called period) for strategic-layer decision-making and a small timescale (called slot) for tactical-layer decision making. Moreover, we design a Lyapunov-based algorithm to solve the tactical layer problem at each time slot, and a Markov approximation algorithm to solve the strategic-layer problem in every time period. Simulation results demonstrate that our proposed solution significantly outperforms existing benchmarks that overlook the coupling between decisions and information realization, achieving up to 56% reduction in total system cost.

Generative Diffusion Model (GDM)-based AI-Generated Content (AIGC) services are rapidly emerging as powerful solutions for creating high-quality, personalized content across various domains, playing an essential role in shaping the future network landscapes. However, the traditional cloud-based implementation of AIGC services faces substantial challenges due to the heterogeneity of GDMs as well as their computation-intensive nature and the low-latency requirement. In this work, we investigate a Mobile Edge Computing (MEC)-empowered heterogeneous AIGC service scenario, where User Equipments (UEs) and Base Stations (BSs) deploy lightweight and heavyweight GDMs, respectively, to deliver different levels of AIGC services at the network edge. Specifically, when initiating an AIGC task, each UE can choose to process the task locally using lightweight GDMs, or offload and process the task on a BS using heavyweight GDMs. In such a scenario, we focus on the joint optimization of offloading, scheduling, and inferencing, aiming to minimize task delay and energy consumption, while maximizing content quality. To address the problem in an online and decentralized manner, we develop a Multi-Agent Proximal Policy Optimization (MAPPO)-based deep reinforcement learning algorithm in a centralized training and decentralized execution framework. Simulation results show that our proposed algorithm outperforms existing intelligent benchmarks, with the performance gains ranging from 5.1% to 13.9%.

Low Earth Orbit (LEO) observation satellite systems play a critical role in a variety of applications, including environmental monitoring, urban planning, and national security. However, transmitting the data collected by numerous observation satellites to Earth remains a significant challenge. One promising approach to improve data transmission efficiency is to utilize LEO communication satellites as data relays. Existing researches in this area primarily focus on the inter-satellite communication scheduling, often neglecting the importance of satellite observation scheduling, which limits the potential performance gain. In this work, we investigate the joint optimization of observation and transmission scheduling in a satellite network with resource-constrained LEO satellites, taking into account the heterogeneity of observation missions and the dynamics of network topology. Specifically, we first introduce a Time-Expanded Graph (TEG) model to effectively represent dynamic network topology and satellite resource constraints. Based on this model, we formulate a network flow problem that incorporates both mission-specific characteristics and satellite energy costs. To reduce the solution complexity in large-scale networks, we propose a novel low-complexity Augmented Lagrangian-based Distributed Parallel Splitting (ALDPS) algorithm. Simulation results show that our proposed algorithm can improve the network utility by 8.3% to 28.1% compared to baseline methods.

AI-Generated Content (AIGC) based on Generative Diffusion Models (GDMs) is revolutionizing content creation and promoting substantial advancements in domains like autonomous driving and robotics. Leveraging progress in Mobile Edge Computing (MEC) and model compression techniques, GDMs are increasingly being deployed on Edge Servers (ESs) and User Equipments (UEs), which typically face resource limitations. In such MEC-enabled scenarios, designing an efficient inference scheme for GDMs still remains a significant challenge, due to the resource constraints on ESs and UEs as well as the personalized demands of AIGC users. In this work, we propose a novel hybrid inference scheme, which consists of two stages: public prompt generation and common-to-personalized inference. In the first stage, a Large Language Model (LLM) is adopted to generate public prompts derived from the common features of users' personal prompts. In the second stage, a common inference phase based on public prompts is first executed for all users (to produce common intermediate results), and then a personalized inference phase based on each user's personal prompts is performed for each individual user (to generate final contents). Clearly, by introducing the common inference phase, the total inference steps can be significantly reduced. In such a scheme, we further study a hybrid inference optimization problem to optimize both common and personalized inference steps, aiming to maximize the total Quality of Service (QoS), while minimizing delay and energy consumption. Simulation results show that our proposed scheme significantly outperforms existing benchmarks, with the performance gains ranging from 12.6 % to 102.2 %.

Multi-UAV cooperative search (MCS) can significantly enhance the efficiency and effectiveness of search by enabling multiple unmanned aerial vehicles (UAVs) to collaborate in conducting search missions. Thus, it has played a vital role in various applications, such as surveillance, target detection, and information gathering. While existing works in this field mainly focused on a single UAV layer, in this work we consider a multilayered aerial computing network (MACN) scenario, which consists of a low-altitude platform (LAP) layer with multiple high-flexibility and low-capacity UAVs (called LUAVs) and a high-altitude platform (HAP) layer with one low-flexibility and high-capacity UAV (called HUAV). In such a scenario, We focus on the joint optimization of flying trajectories, computation offloading, and resource allocation, aiming at minimizing the uncertainty of search probability map (SPM), and meanwhile maximizing the number of target discovery and coverage rate. The problem is challenging due to the co-existence of discrete and continuous decision variables, as well as the fast and randomly changing of wireless environment. To solve the problem in an online and distributed manner, we propose a multiagent deep reinforcement learning (MADRL) approach based on the parameter sharing and action mask (PSAM), called PSAMMA, where the state-action-reward-state-action (SARSA) method is leveraged to determine the discrete flying and offloading decisions. Experiment results show that 1) the proposed PSAMMA algorithm outperforms existing algorithms in the literature, and can increase the average utility by 9.89%–31.15% and 2) we evaluate the search performance by analyzing the average uncertainty, target rate, and coverage rate under different parameter settings.

Artificial Intelligence-Generated Content (AIGC) has emerged as a transformative paradigm, enabling the autonomous creation of diverse content. By offloading model inference tasks to the network edge that is closer to mobile users (MUs), Mobile Edge Computing (MEC) has the potential to significantly enhance the performance of AIGC services. In practice, however, it is challenging to optimally manage MEC-empowered AIGC services, due to the lack of well-defined AIGC-specific metrics, as well as the dynamic workload and computation-intensive nature of AIGC services. In this paper, we first define a novel AIGC metric based on extensive real data experiments, and then study thejoint task offloading and resource allocationproblem in a generic MEC-empowered AIGC network, where MUs can offload model inference tasks to local or remote Base Stations (BSs), aiming at maximizing their Quality of Experience (QoE). The problem is challenging due to the fast and randomly changing of environments, as well as the necessity for real-time, asynchronous decision-making. To tackle these challenges, we propose two deep reinforcement learning algorithms based on the Proximal Policy Optimization (PPO) framework:Single-Layer PPO (SL-PPO)andMulti-Layer PPO (ML-PPO), designed for slow-changing and fast-changing environments, respectively. In the SL-PPO algorithm, both task offloading and resource allocation decisions are made simultaneously when tasks arrive. In the ML-PPO algorithm, the task offloading decision is made immediately when tasks arrive, while the resource allocation decision is deferred until tasks are scheduled for processing or transmission in the corresponding queues. Simulation results show that (i) both algorithms outperform existing methods in the literature, and can increase the average utility by up to 47% and 48.8%; (ii) both algorithms can effectively manage the trade-off between latency and energy consumption.

Multi-UAV Cooperative Search (MCS) can significantly enhance the efficiency and effectiveness of search by enabling multiple unmanned aerial vehicles (UAVs) to collaborate in conducting search missions. Thus, it has played a vital role in various applications, such as surveillance, target detection, and information gathering. While existing works in this field mainly focused on a single UAV layer, in this work we consider a Multi-layered Aerial Computing Network (MACN) scenario, which consists of a Low-Altitude Platform (LAP) layer with multiple high-flexibility and low-capacity UAVs (called LUAVs) and a High-Altitude Platform (HAP) layer with one low-flexibility and high-capacity UAV (called HUAV). In such a scenario, We focus on the joint optimization of flying trajectories, computation offloading, and resource allocation, aiming at minimizing the uncertainty of Search Probability Map (SPM). The problem is challenging due to the co-existence of discrete and continuous decision variables, as well as the fast and randomly changing of wireless environment. To solve the problem in an online and distributed manner, we propose a Multi-Agent Deep Reinforcement Learning (MADRL) approach based on Parameter Sharing and Action Mask (PSAM), called PSAMMA, where the State-Action-Reward-State-Action (SARSA) method is leveraged to determine the discrete flying and offloading decisions. Experiment results show that the proposed PSAMMA algorithm outperforms existing methods in terms of the average SPM uncertainty, the target discovery rate, and the coverage rate.

Unmanned Aerial Vehicles (UAVs), with their high levels of flexibility and maneuverability, can greatly enhance the capabilities of Mobile Edge Computing (MEC) by acting as edge computing servers. In practice, however, it is often challenging to jointly optimize the flying trajectories of UAVs and the offloading decisions of tasks, due to the fast and randomly changing of physical environments. In this work, we investigate an UAV-enable MEC network with the assistance of Digital Twin (DT), where a DT layer is introduced to simulate the Physical Entity (PE) layer, generate different strategies, and evaluate their performances. Specifically, we formulate a joint flying trajectories, task offloading, and resource allocation problem on the DT layer, aiming at minimizing both task delay and energy consumption, under the maximum tolerated delay and resource constraints. To solve the problem in an online distributed manner and implement the derived strategies on the real PE layer, we propose a hierarchical learning approach, which consists of a Deep Reinforcement Learning (DRL) module and a Constrained Optimization (CO) module. First, the DRL module determines the UAVs' flying trajectories. Then, the CO module determines the MDs' task offloading decisions and the associated resource allocations, given the UAV s' flying decisions. Finally, the outputs of both modules are combined together to train the DRL module by using the Deep Deterministic Policy Gradient (DDPG) method. Experiment results show that our proposed DT-assisted scheme outperforms existing benchmark schemes in terms of both task delay and energy cost.

Mobile-edge computing (MEC) is a promising computing scheme to support computation-intensive AI applications in vehicular networks, by enabling vehicles to offload computation tasks to edge computing servers deployed on road side units (RSUs) that approximate to them. In this work, we consider an MEC-enabled vehicular edge network (VEN), where each vehicle can offload tasks to edge/cloud computing servers via vehicle-to-infrastructure (V2I) links or to other end-vehicles via vehicle-to-vehicle (V2V) links. In such acloud-edge–endcollaborative offloading scenario, we focus on the joint task offloading, scheduling, and resource allocation problem for vehicles, which is challenging due to the online and asynchronous decision-making requirement for each task. To solve the problem, we propose aMultilayer deep reinforcement learning(DRL)-based approach, where each vehicle constructs and trains three modules to make different layers’ decisions: 1)Offloading Module(first layer), determining whether to offload each task, by using the dueling and double deepQ-network (D3QN) framework; 2)Scheduling Module(second layer), determining where and how to offload each task in the offloading queues, together with the transmission power, by using the parameterized deepQ-network (PDQN) framework; and 3)Computing Module(third layer), determining how much computing resource to be allocated for each task in the computation queues, by using classic optimization techniques. We provide the detailed algorithm design and perform extensive simulations to evaluate its performance. Simulation results show that our proposed algorithm outperforms the existing algorithms in the literature, and can reduce the average cost by 25.86%–72.51% and increase the average satisfaction rate by 3.48%–90.53%.

Mobile Edge Computing (MEC) is emerging as a promising computing scheme to support AI-enabled applications in vehicular networks, via offloading some tasks to edge servers deployed on Road Side Units (RSUs) that approximates to vehicles. In this work, we consider a general vehicular edge network (VEN), where each vehicle can offload tasks to edge servers or cloud server via a vehicle-to-infrastructure (V2I) transmission link, or to other vehicles via a vehicle-to-vehicle (V2V) transmission link. To characterize different task flows in different transmission links or computing servers, we introduce a V2V transmission queue, a V2V transmission queue, and a local computation queue for each vehicle, and an edge computation queue for each edge server. In such a queue-based VEN, we focus on the joint task offloading and scheduling problem for vehicles, which consists of (i) offloading problem, i.e., whether to offload tasks, and (ii) scheduling problem, i.e., where and how to offload tasks. The problem is challenging due to the online and asynchronous offloading and scheduling decisions for each task. We propose a Multi-layer Deep Reinforcement Learning (DRL) approach, where each vehicle trains three neural networks (called agents) to make different layers' decisions: (i) offloading agent, determining whether to offload each task when tasks arrive, and (ii) V2I and V2V scheduling agents, determining where and how to offloading each task in V2I and V2V transmission queues, respectively. We provide the detailed algorithm design of each agent by using the Double Deep Q-Network (DDQN) approach. Simulation results show that our proposed multi-layer DRL approach outperforms the existing baseline approaches in terms of both the cost performance and the convergence speed.

Mobile Edge Computing (MEC) is a promising approach for enhancing the quality-of-service (QoS) of AI-enabled applications in the B5G/6G era, via providing computation services at the network edge that approximate end-users. In this work, we focus on the joint optimization of edge server (ES) deployment, service placement, and computation task offloading under stochastic information scenario. In traditional solutions, these decisions are often treated equally without considering differences in the information realization. In practice, however, the ES deployment decision needs to be made in advance before the complete information is realized, while the service placement and computation task offloading decisions can be made after the complete information is realized. To capture the time coupling between different decisions and information realizations, we formulate a two-layer stochastic programming (SP) problem, which consists of a strategic-layer decision for ES deployment, and a tactical-layer decision for service placement and computation task offloading. The strategic-layer decision will be made based on the stochastic information (i.e., before the complete information is realized), while the tactical-layer decision will be made based on every information realization. The problem is very challenging due to the large number of information realizations and the corresponding tactical-layer decisions. To solve the problem effectively, we propose a Sample Average Approximate (SAA) method to approximate the optimal solution, which involves generating a large number of randomly sampled information scenarios and using their averages to estimate the expected value of the objective function. Numerical simulations show that our proposed SP approach outperforms the traditional solutions that do not consider the coupling between decisions and information realizations. Moreover, compared with the ideal benchmark solution that assumes complete information, our proposed SP approach only results in a small performance degradation of 1.03%$\sim$6.26%.

Mobile edge computing (MEC) is a promising approach to reduce the network traffic load and alleviate the back-haul congestion by pushing computation down to the network edge (e.g., base stations) that are close to the origin of data. However, when many mobile devices (MDs) offload tasks to a base station (BS) in a dynamic and stochastic environment (e.g., with time-varying wireless channels and uncertain task models), it is often challenging for MDs to make offloading decisions in decentralized manner. In this work, we consider a collaborative MEC scenario, where an MD can offload its task to the associated BS or to other BSs through the associated BS. In such a scenario, we study the joint computation offloading and resource allocation problem, aiming at minimizing the expected long-term delay, taking the energy consumption constraint into consideration. The problem is challenging due to time-varying system and distributed decisions. To solve the problem in an online and decentralized manner, we propose a deep reinforcement learning (DRL) based distributed online algorithm. By incorporating the double deep Q network and dueling deep Q network technique, the proposed algorithm can improve the performance of the whole system significantly. Simulation results show that the proposed DRL-based algorithm outperforms baseline methods and can reduce the average delay of tasks by 76.4%-91.2%.