pith. sign in

arxiv: 2605.22277 · v1 · pith:FXGYLRSXnew · submitted 2026-05-21 · 💻 cs.GT

Joint Communication and Computation Scheduling for MEC-enabled AIGC Services: A Game-Theoretic Stochastic Learning Approach

Huaizhe Liu , Xinyi Zhuang , Jiaqi Wu , Yuan Luo , Bin Cao , Lin Gao This is my paper

Pith reviewed 2026-05-22 02:27 UTC · model grok-4.3

classification 💻 cs.GT
keywords mobile edge computingAIGCpotential gameNash equilibriumstochastic learningcomputation offloadingdiffusion modelsgame theory
0
0 comments X

The pith

Users in edge networks can coordinate access points, servers, and inference steps through a potential game to cut AI content creation delays with only local information.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper models the joint selection of wireless access points, edge servers, and diffusion-model inference steps by mobile users as a game that each user plays to minimize its own service completion time while respecting accuracy limits. It proves this JCACO game admits a potential function under both full information and stochastic information about network conditions, which guarantees that a Nash equilibrium always exists. The authors then introduce a fully distributed Multi-Agent Stochastic Learning algorithm that updates each user's choices using only its own observations and provably converges to the equilibrium. Simulations show the resulting schedules reduce completion time relative to standard benchmarks while still meeting the accuracy targets.

Core claim

The JCACO game in which each user selects a serving access point, edge server, and number of inference steps is a potential game under both complete and stochastic information settings. This property ensures the existence of a Nash equilibrium in both cases. A distributed Multi-Agent Stochastic Learning algorithm converges to that equilibrium with strict performance guarantees and requires neither knowledge of other users' strategies nor global network state.

What carries the argument

The potential function of the JCACO game, which strictly decreases with any unilateral improvement in a user's strategy and thereby guarantees convergence to equilibrium through local updates even when channel and load information is stochastic.

If this is right

  • A Nash equilibrium exists even when users observe only noisy or delayed network state.
  • The MASL algorithm reaches equilibrium without exchanging strategy information among users.
  • Service completion time decreases while accuracy constraints are met in time-varying wireless environments.
  • The scheme adapts automatically to changes in user demand or server load because updates are purely local.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same potential-game structure might apply to other edge AI workloads that trade off latency against output quality.
  • Hardware-in-the-loop experiments with measured wireless traces would test whether the stochastic-information model holds in practice.
  • Larger user populations may require tuning of the learning step size to keep convergence time acceptable.

Load-bearing premise

That users' choices of access point, edge server, and inference steps can be modeled so the resulting interaction satisfies the potential-game property under stochastic network information.

What would settle it

A dynamic network simulation in which the MASL updates fail to produce a strategy profile whose service completion time is stable and no higher than the benchmarks while accuracy constraints remain satisfied.

Figures

Figures reproduced from arXiv: 2605.22277 by Bin Cao, Huaizhe Liu, Jiaqi Wu, Lin Gao, Xinyi Zhuang, Yuan Luo.

Figure 1
Figure 1. Figure 1: An example of MEC-enabled GDM-based AIGC Network. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Dynamics of UE’s service completion time. [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Impact of different of learning rate settings on MASL convergence. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Total service time vs. the number of APs. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Total service time vs. wireless bandwidth of APs. [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 2
Figure 2. Figure 2: This behavior indicates that the MASL algorithms [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 8
Figure 8. Figure 8: Total service time vs. TFLOPs per inference step [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Total service time vs. the number of UEs. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

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 a Joint 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 a potential game under both complete and stochastic information settings, ensuring the existence of Nash Equilibrium (NE) in both cases. To derive the NE efficiently, we develop a distributed Multi-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 using Ordinary 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.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper formulates a Joint Communication Association and Computation Offloading (JCACO) game for MEC-enabled AIGC networks in which each UE selects a serving AP, ES, and number of inference steps to minimize service completion time subject to accuracy constraints. It claims to prove that the game is an exact potential game (hence admits a Nash equilibrium) under both complete-information and stochastic-information settings, proposes a fully distributed Multi-Agent Stochastic Learning (MASL) algorithm that converges to the NE, and supplies an ODE-based convergence analysis together with simulation results showing reduced completion time relative to benchmarks.

Significance. If the potential-game property and ODE convergence hold under the stated stochastic model, the work supplies a theoretically grounded, fully distributed method for latency-accuracy trade-offs in dynamic edge AIGC settings. The explicit ODE analysis and the absence of any requirement for other players’ strategies or global state are genuine strengths that distinguish the contribution from conventional best-response schemes.

major comments (2)
  1. [JCACO game formulation and potential-game proof] The central claim that the JCACO game remains an exact potential game under stochastic information (abstract and theoretical sections) is load-bearing for both NE existence and the applicability of MASL. The expected utilities incorporate user-specific stochastic wireless channels and server loads; the manuscript must explicitly construct the potential function or verify that any unilateral deviation changes the expected utility by exactly the same amount as the potential, rather than inheriting the property from the complete-information case without additional argument.
  2. [Simulation results] Table or simulation section reporting performance gains: the claimed improvements over benchmarks must be shown to arise from the learned NE rather than from post-hoc tuning of the inference-step count or accuracy threshold; otherwise the cross-method comparison risks being driven by modeling choices rather than by the game-theoretic solution.
minor comments (2)
  1. [System model] Notation for the stochastic information setting (e.g., how the expectation is taken over channel and load realizations) should be introduced once and used consistently; currently the transition from complete to stochastic information is abrupt.
  2. [Abstract] The abstract states “strict performance guarantees”; the manuscript should state explicitly what these guarantees are (e.g., finite-time convergence bounds or regret bounds) rather than leaving them implicit in the ODE analysis.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments. We address each major comment below and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [JCACO game formulation and potential-game proof] The central claim that the JCACO game remains an exact potential game under stochastic information (abstract and theoretical sections) is load-bearing for both NE existence and the applicability of MASL. The expected utilities incorporate user-specific stochastic wireless channels and server loads; the manuscript must explicitly construct the potential function or verify that any unilateral deviation changes the expected utility by exactly the same amount as the potential, rather than inheriting the property from the complete-information case without additional argument.

    Authors: We agree that an explicit construction strengthens the stochastic-information case. While the manuscript derives the stochastic potential by taking the expectation of the complete-information potential (leveraging linearity of expectation over the independent channel and load random variables), we will add a dedicated paragraph in the theoretical section that directly defines the stochastic potential function Φ and verifies that, for any unilateral deviation, the change in a player's expected utility equals the change in Φ. This verification uses the structure of the service completion time expression and holds under the stated stochastic model. revision: yes

  2. Referee: [Simulation results] Table or simulation section reporting performance gains: the claimed improvements over benchmarks must be shown to arise from the learned NE rather than from post-hoc tuning of the inference-step count or accuracy threshold; otherwise the cross-method comparison risks being driven by modeling choices rather than by the game-theoretic solution.

    Authors: We clarify that inference steps and accuracy thresholds are endogenous to each UE's strategy in the JCACO game and are subject to the same hard accuracy constraints for MASL and all benchmarks. To make this explicit and rule out post-hoc tuning, we will revise the simulation section to include an additional figure and table that compare the learned NE against fixed inference-step policies (minimum, maximum, and average steps) while enforcing identical accuracy constraints. This will isolate the contribution of the equilibrium strategy selection. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on explicit potential function construction and independent ODE convergence analysis

full rationale

The paper defines the JCACO game with utilities based on service completion time and accuracy constraints, then proves the potential-game property by exhibiting a potential function that matches unilateral deviation payoffs under both complete and stochastic information. This construction is shown directly rather than assumed or fitted. The MASL algorithm's convergence is grounded in a separate ODE analysis that does not reference the fitted simulation outcomes or reduce to the input strategy space by definition. No self-citation chains, ansatz smuggling, or renaming of known results appear as load-bearing steps. The modeling choice for stochastic utilities is an assumption that is then verified through the potential-function proof, not circularly presupposed.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract; the central claims rest on standard potential-game theory and stochastic approximation. No explicit free parameters or invented entities are named. Specific simulation parameters and exact payoff functions are not provided.

axioms (2)
  • domain assumption The JCACO interaction among UEs selecting AP, ES, and inference steps forms a potential game under both complete and stochastic information.
    Invoked to guarantee Nash equilibrium existence; stated as proven in the abstract.
  • standard math Ordinary differential equations can be used to analyze convergence of the multi-agent stochastic learning dynamics.
    Used for the strict theoretical convergence analysis mentioned in the abstract.

pith-pipeline@v0.9.0 · 5844 in / 1373 out tokens · 53437 ms · 2026-05-22T02:27:01.370008+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages

  1. [1]

    Joint Communication and Computation Scheduling for MEC-Enabled AIGC Services Based on Generative Diffusion Model,

    H. Liu, J. Wu, X. Zhuang, H. Wu and L. Gao, “Joint Communication and Computation Scheduling for MEC-Enabled AIGC Services Based on Generative Diffusion Model,” inProc. WiOpt, Korea, Oct. 2024

    work page 2024

  2. [2]

    Unleashing the Power of Edge-Cloud Generative AI in Mobile Networks: A Survey of AIGC Services,

    M. Xu, H. Du, D. Niyato, J. Kang, and J. Xiong, et al., “Unleashing the Power of Edge-Cloud Generative AI in Mobile Networks: A Survey of AIGC Services,”IEEE Commun. Surveys Tuts., vol. 26, no. 2, pp. 1127-1170, Secondquarter 2024

    work page 2024

  3. [3]

    Diffusion Models in Vision: A Survey,

    F. -A. Croitoru, V . Hondru, R. T. Ionescu and M. Shah, “Diffusion Models in Vision: A Survey,”IEEE Trans. Pattern Anal. Mach. Intell., pp. 10850-10869, Sept. 2023

    work page 2023

  4. [4]

    Come-Closer-Diffuse-Faster: Accel- erating Conditional Diffusion Models for Inverse Problems Through Stochastic Contraction,

    H. Chung, B. Sim, and J. C. Ye, “Come-Closer-Diffuse-Faster: Accel- erating Conditional Diffusion Models for Inverse Problems Through Stochastic Contraction,”Proc. IEEE CVPR, 2022, pp. 12 413–12 422

    work page 2022

  5. [5]

    High-Resolution Image Synthesis with Latent Diffusion Models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-Resolution Image Synthesis with Latent Diffusion Models,” Proc. IEEE CVPR, 2022, pp. 10 684–10 695

    work page 2022

  6. [6]

    Versatile Diffusion: Text, Images and Variations All in One Diffusion Model,

    X. Xu, Z. Wang, E. Zhang, K. Wang and H. Shi, “Versatile Diffusion: Text, Images and Variations All in One Diffusion Model,”Proc. IEEE ICCV, 2023, pp. 7720-7731

    work page 2023

  7. [7]

    Monetizing Edge Service in Mobile Internet Ecosystem,

    Z. Wang, L. Gao, T. Wang, and J. Luo, “Monetizing Edge Service in Mobile Internet Ecosystem,”IEEE Trans. Mob. Comput., vol.21, no.5, pp.1751-1765, 2022

    work page 2022

  8. [8]

    A Truthful Incentive Mecha- nism for Movement-Aware Task Offloading in Crowdsourced Mobile Edge Computing Systems,

    C. Jiang, Z. Luo, L. Gao, and J. Li, “A Truthful Incentive Mecha- nism for Movement-Aware Task Offloading in Crowdsourced Mobile Edge Computing Systems,”IEEE Internet Things J., vol.11, no.10, pp.18292-18305, 2024

    work page 2024

  9. [9]

    Price Competition in Multi- Server Edge Computing Networks under SAA and SIQ Models,

    Z. Chen, Q. Ma, L. Gao, and X. Chen, “Price Competition in Multi- Server Edge Computing Networks under SAA and SIQ Models,”IEEE Trans. Mob. Comput., vol.23, no.1, pp.754-768, 2024

    work page 2024

  10. [10]

    Diffusion-based Reinforcement Learning for Edge-enabled AI-Generated Content Services,

    H. Du, Z. Li, N. Dusit,et al.,“Diffusion-based Reinforcement Learning for Edge-enabled AI-Generated Content Services,”IEEE Trans. Mob. Comput., pp. 1-16, Jan. 2024

    work page 2024

  11. [11]

    Beyond the Cloud: Edge Inference for Generative Large Language Models in Wireless Networks,

    X. Zhang, J. Nie, Y . Huang, G. chang, Z. Xiong, J. Liu, D. Niyato, and X. Shen, “Beyond the Cloud: Edge Inference for Generative Large Language Models in Wireless Networks,”IEEE Trans. Wireless Commun., vol.24, no.1, pp.643-658, 2025

    work page 2025

  12. [12]

    Characterizing and Scheduling of Diffusion Process for Text-to-Image Generation in Edge Networks,

    S. Gao, P. Yang, Y . Kong, F. Lyu, and N. Zhang, “Characterizing and Scheduling of Diffusion Process for Text-to-Image Generation in Edge Networks,”IEEE Trans. Mob. Comput., early access, 2025

    work page 2025

  13. [13]

    Efficient Multi-user Offloading of Personalized Diffusion Models: A DRL- Convex Hybrid Solution,

    W. Yang, Z. Xiong, S. Guo, S. Mao, D. Kim, and M. Debbah,“Efficient Multi-user Offloading of Personalized Diffusion Models: A DRL- Convex Hybrid Solution,”IEEE Trans. Mob. Comput., 2025

    work page 2025

  14. [14]

    Exploring Collaborative Diffusion Model Inferring for AIGC-Enabled Edge Services,

    W. Feng, R. Zhang, Y . Zhu,et al., “Exploring Collaborative Diffusion Model Inferring for AIGC-Enabled Edge Services,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 2, pp. 946-960, 2025

    work page 2025

  15. [15]

    QoS- Aware Multi-AIGC Service Orchestration at Edges: An Attention- Diffusion-Aided DRL Method,

    Y . Liu, S. Li, X. Lin, X. Chen, G. Li, Y . Liu, B. Liao, and J. Li, “QoS- Aware Multi-AIGC Service Orchestration at Edges: An Attention- Diffusion-Aided DRL Method,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 2, pp. 1078-1090, Apr. 2025

    work page 2025

  16. [16]

    Dynamic Trajectory Optimization and Power Control for Hierarchical UA V Swarms in 6G Aerial Access Network,

    Z. Jia, J. He, L. He, M. Sheng, J. Liu, Q. Wu, and Z. Han, “Dynamic Trajectory Optimization and Power Control for Hierarchical UA V Swarms in 6G Aerial Access Network,”IEEE Trans. Wireless Commu., vol. 25, pp. 3349-3362, 2026

    work page 2026

  17. [17]

    Hierarchical Low-Altitude Wireless Network Empowered Air Traffic Management,

    Z. Jia, J. He, Y . Cui, Q. Zhu, L. Yuan, F. Zhou, Q. Wu, D. Niyato, and Z. Han, “Hierarchical Low-Altitude Wireless Network Empowered Air Traffic Management,”IEEE Commun. Mag., early access, 2026

    work page 2026

  18. [18]

    A Blockchain-Enabled Cold Start Aggregation Scheme for Federated Reinforcement Learning-Based Task Offloading in Zero Trust LEO Satellite Networks,

    B. Mao, Y . Liu, Z. Wei,et al., “A Blockchain-Enabled Cold Start Aggregation Scheme for Federated Reinforcement Learning-Based Task Offloading in Zero Trust LEO Satellite Networks,”IEEE J. Sel. Areas Commun., vol. 43, pp. 2172-2182, 2025

    work page 2025

  19. [19]

    On an Intelligent Collaborative Computation Strategy for Low Earth Orbit Satellite Edge Computing Networks,

    B. Mao, Z. Xia, Y . Yin,et al., “On an Intelligent Collaborative Computation Strategy for Low Earth Orbit Satellite Edge Computing Networks,”IEEE Internet. Things J., early access, 2026

    work page 2026

  20. [20]

    AIGC-Enhanced Hybrid Content Caching in Wireless Networks,

    D. Xu, L. Duan, and H. Zhu, “AIGC-Enhanced Hybrid Content Caching in Wireless Networks,”IEEE Trans. Wireless Commun., 2025

    work page 2025

  21. [21]

    DRL-Enabled Computation Offloading for AIGC Services in IoIT-assisted Edge Computing Networks,

    X. Zhang, S. Li,et al., “DRL-Enabled Computation Offloading for AIGC Services in IoIT-assisted Edge Computing Networks,”IEEE Internet Things J., vol. 12, no. 9, pp. 12829-12844, 2024

    work page 2024

  22. [22]

    Joint Optimization of Model Inferencing and Task Offloading for MEC-Empowered Large Vision Model Services,

    X. Zhuang, J. Wu, H. Wu, T. Zhang, and L. Gao, “Joint Optimization of Model Inferencing and Task Offloading for MEC-Empowered Large Vision Model Services,” inProc. IEEE Int. Conf. Comput. Commun.(INFOCOM), London, England, May, 2025

    work page 2025

  23. [23]

    QoE-Aware Offloading and Resource Allocation for MEC-Empowered AIGC Services,

    J. Wu, X. Zhuang, M. Tang, and L. Gao, “QoE-Aware Offloading and Resource Allocation for MEC-Empowered AIGC Services,”IEEE Trans. Mob. Comput., early access, 2025

    work page 2025

  24. [24]

    Computation Energy Efficiency Maximization for NOMA-Based and Wireless-Powered Mobile Edge Computing With Backscatter Communication,

    J. Du, H. Wu, M. Xu, and R. Buyya, “Computation Energy Efficiency Maximization for NOMA-Based and Wireless-Powered Mobile Edge Computing With Backscatter Communication,”IEEE Trans. Mob. Comput., vol. 23, no. 6, pp. 6954-6970, 2024

    work page 2024

  25. [25]

    Caching-Assisted Collaborative Task Offloading for Vehicular Edge Computing: A Deep Reinforce- ment Learning-Based Approach,

    C. Tang, S. Wang, H. Wu, and R. Li, “Caching-Assisted Collaborative Task Offloading for Vehicular Edge Computing: A Deep Reinforce- ment Learning-Based Approach,”IEEE Trans. Mob. Comput., early access, 2026

    work page 2026

  26. [26]

    Accelerating Mobile Edge Generation (MEG) by Constrained Learning,

    X. Xu, Y . Liu, X. Mu, H. Xing, and A. Nallanathan, “Accelerating Mobile Edge Generation (MEG) by Constrained Learning,”IEEE Trans. Cogn. Commun. Netw., early access, 2025

    work page 2025

  27. [27]

    Online AI-Generated Content Request Scheduling with Deep Reinforcement Learning,

    C. Feng, Y . Zheng, and Y . Xu, “Online AI-Generated Content Request Scheduling with Deep Reinforcement Learning,” in Proc. IEEE Int. Conf. Comput. Commun.(INFOCOM WKSHPS), Vancouver, BC, Canada, 2024, pp. 1-6

    work page 2024

  28. [28]

    Contract-Inspired Contest Theory for Controllable Image Generation in Mobile Edge Metaverse,

    G. Liu, H. Du, J. Wang, D. Niyato, and D. Kim, “Contract-Inspired Contest Theory for Controllable Image Generation in Mobile Edge Metaverse,”IEEE Trans. Mob. Comput., early access, 2025

    work page 2025

  29. [29]

    Diffusion-Model-Based Incentive Mechanism With Prospect Theory for Edge AIGC Services in 6G IoT,

    J. Wen, J. Nie, Y . Zhong, C. Yi, X. Li, J. Jin, Y . Zhang, and D. Niyato, “Diffusion-Model-Based Incentive Mechanism With Prospect Theory for Edge AIGC Services in 6G IoT,”IEEE Internet Things J., vol. 11, no. 21, pp. 34187-34201, Nov. 2024

    work page 2024

  30. [30]

    Edge AI Inference as a Service via Dynamic Resources from Repeated Auctions,

    M. Ji, H. Zhao, L. Jiao, S. Zhang, X. Li, Z. Qian, and B. Ye, “Edge AI Inference as a Service via Dynamic Resources from Repeated Auctions,”IEEE Trans. Mob. Comput., early access, 2025

    work page 2025

  31. [31]

    Optimiz- ing AIGC Services by Prompt Engineering and Edge Computing: A 16 Generative Diffusion Model-Based Contract Theory Approach,

    D. Ye, S. Cai, H. Du, J. Kang, Y . Liu, R. Yu, and D. Niyato, “Optimiz- ing AIGC Services by Prompt Engineering and Edge Computing: A 16 Generative Diffusion Model-Based Contract Theory Approach,”IEEE Trans. Veh. Technol., vol. 74, no. 1, pp. 571-586, Jan. 2025

    work page 2025

  32. [32]

    Resource- Efficient Generative Mobile Edge Networks in 6G Era: Fundamentals, Framework and Case Study,

    B. Lai, J. Wen, J. Kang, H. Du, J. Nie, and C. Yi, “Resource- Efficient Generative Mobile Edge Networks in 6G Era: Fundamentals, Framework and Case Study,”IEEE Wireless Commun., vol. 31, no. 4, pp. 66-74, Aug. 2024

    work page 2024

  33. [33]

    A Secure and Intelligent Data Sharing Scheme for UA V-Assisted Disaster Rescue,

    Y . Wang, Z. Su, Q. Xu, R. Li, T. Luan, and, P. Wang, “A Secure and Intelligent Data Sharing Scheme for UA V-Assisted Disaster Rescue,” IEEE/ACM Trans. Netw., vol. 31, no. 6, pp. 2422-2438, 2023

    work page 2023

  34. [34]

    Optimizing AIGC Services Using Learning-Based Stackelberg Game in Vehicular Metaverses,

    B. Lai, X. Luo, J. Kang, X. Gao, Z. Yang, D. Niyato, and S. Mao, “Optimizing AIGC Services Using Learning-Based Stackelberg Game in Vehicular Metaverses,”IEEE Trans. Veh. Technol., vol. 74, no. 7, pp. 11472-11477, 2025

    work page 2025

  35. [35]

    Federated MADDPG-Based Collabora- tive Scheduling Strategy in Vehicular Edge Computing,

    S. Lei, H. Tang, C. Li,et al., “Federated MADDPG-Based Collabora- tive Scheduling Strategy in Vehicular Edge Computing,”IEEE Trans. Mob. Comput., vol. 25, no. 1, pp. 54-66, 2026

    work page 2026

  36. [36]

    Optimizing Resource Allocation for Multi-modal Semantic Communication in Mobile AIGC Networks: A Diffusion- based Game Approach,

    J. Liu, M. Xiao, J. Wen, J. Kang, R. Zhang, T. Zhang, D. Niyato, W. Zhang, and Y . Liu, “Optimizing Resource Allocation for Multi-modal Semantic Communication in Mobile AIGC Networks: A Diffusion- based Game Approach,”IEEE Trans. Cogn. Commun. Netw., 2025

    work page 2025

  37. [37]

    Opportunistic Spectrum Access in Unknown Dynamic Environment: A Game- Theoretic Stochastic Learning Solution,

    Y . Xu, J. Wang, Q. Wu, A. Anpalagan, and Y . Yao,“Opportunistic Spectrum Access in Unknown Dynamic Environment: A Game- Theoretic Stochastic Learning Solution,”IEEE Trans. Wireless Com- mun., vol.11, no.4, Apr. 2012

    work page 2012

  38. [38]

    Stochastic game- theoretic spectrum access in distributed and dynamic environment,

    J. Zheng, Y . Cai, N. Lu, Y . Xu, and X. Shen, “Stochastic game- theoretic spectrum access in distributed and dynamic environment,” IEEE Trans. Veh. Technol., vol. 64, no. 10, pp. 4807–4820, Oct. 2015

    work page 2015

  39. [39]

    Dynamic Computation Of- floading for Mobile Cloud Computing: A Stochastic Game-Theoretic Approach,

    J. Zheng, Y . Cai, Y . Wu and X. Shen, “Dynamic Computation Of- floading for Mobile Cloud Computing: A Stochastic Game-Theoretic Approach,”IEEE Trans. Mob. Comput., vol. 18, no. 4, 2019

    work page 2019

  40. [40]

    Optimal Pricing Design for Coordi- nated and Uncoordinated IoT Networks,

    N. Ding, L. Gao, and J. Huang, “Optimal Pricing Design for Coordi- nated and Uncoordinated IoT Networks,”IEEE Trans. Mob. Comput., vol.22, no.12, pp.7121-7137, 2023

    work page 2023

  41. [41]

    A Game-Theoretic Analysis of Joint Mobile Edge Caching and Peer Content Sharing,

    C, Jiang, L. Gao, J. Luo, P. Zhou, and J. Li, “A Game-Theoretic Analysis of Joint Mobile Edge Caching and Peer Content Sharing,” IEEE Trans. Net. Sci. Eng., vol.10, no.3, pp.1445-1461, 2023

    work page 2023

  42. [42]

    Economic Analysis of Edge Caching Enabled Mobile Internet Ecosystem,

    C. Jiang, L. Gao, F. Hou, and J. Li, “Economic Analysis of Edge Caching Enabled Mobile Internet Ecosystem,”IEEE Trans. Mob. Comput., Early Access, 2024

    work page 2024

  43. [43]

    Deep reinforcement learning for task offload- ing in mobile edge computing systems,

    M. Tang, and V . Wong, “Deep reinforcement learning for task offload- ing in mobile edge computing systems,”IEEE Trans. Mob. Comput., vol. 21, no. 6, pp. 1985-1997, Jun. 2022

    work page 1985

  44. [44]

    Scheduling Efficiency of Distributed Greedy Scheduling Algorithms in Wireless Networks,

    X. Wu, R. Srikant and J. R. Perkins, “Scheduling Efficiency of Distributed Greedy Scheduling Algorithms in Wireless Networks,” in IEEE Trans. Mob. Comput., vol. 6, no. 6, pp. 595-605, June 2007

    work page 2007

  45. [45]

    Fudenberg and J

    D. Fudenberg and J. Tirole,Game Theory. MIT Press, 1991

    work page 1991

  46. [46]

    Potential games,

    D. Monderer and L. S. Shapley, “Potential games,”Games Econ. Behav., vol. 14, no. 1, pp. 124–143, 1996

    work page 1996

  47. [47]

    Joint strategy fictitious play with inertia for potential games,

    J. Marden, G. Arslan, and J. Shamma, “Joint strategy fictitious play with inertia for potential games,”IEEE Trans. Automat. Control, vol. 54, no. 2, pp. 208–220, Feb. 2009

    work page 2009

  48. [48]

    Decentralized learning of Nash equilibria in multi-person stochastic games with incomplete information,

    P. Sastry, V . Phansalkar, and M. Thathachar, “Decentralized learning of Nash equilibria in multi-person stochastic games with incomplete information,”IEEE Trans. Syst. Man Cybern., vol. 24, no. 5, pp. 769–777, May 1994

    work page 1994

  49. [49]

    Networks of learningautomata tech- niques for online stochastic optimization

    M. Thathachar and P. Sastry,“Networks of learningautomata tech- niques for online stochastic optimization”, Bangalore, India, Kluwer Academic, 2004

    work page 2004

  50. [50]

    H. J. Kushner, Approximation and Weak Convergence Methods for Random Processes, MIT Press, Cambridge, 1984. Huaizhe Liu(Graduate Student Member, IEEE) is currently pursuing the Ph.D. degree with the School of Information Science and Technology, Harbin In- stitute of Technology, Shenzhen, China. He received the M.S. degree in Information and Communication ...

    work page 1984

  51. [51]

    Case1: Active UEi∈ N A changes its computation offloading strategy while keeps the inference steps un- changed

    work page

  52. [52]

    Case2: Active UEi∈ N A change its inference steps while keeps the computation offloading strategy un- changed

    work page

  53. [53]

    Case1: Suppose that an active UEi∈ N A updates its decisiony i to y′ i (i.e.,y ik = 1→y ik′ = 1), that is, UEiswitches from an ESkto a new onek ′

    Case3: Active UEi∈ N A changes both the computation offloading strategy and the inference steps. Case1: Suppose that an active UEi∈ N A updates its decisiony i to y′ i (i.e.,y ik = 1→y ik′ = 1), that is, UEiswitches from an ESkto a new onek ′. LetYandY ′ = (yi′,Y −i)denote the strategy profiles before and after the strategy update of UEi, respectively. It...

    work page

  54. [54]

    That is, when˜φ≥ ϵ√ 2, we have:sgn(∆ T Acc i (X→X ′)) = sgn(∆Φ(X→X ′)), which implies that the communication association game is a potential game with expected potential function Φ(X) =PM m=1 ˜φLm(X), for any˜φ≥ ϵ√ 2. D. Proof for Theorem 4 Theorem 4.The gameG S2 is a stochastic potential game with the expected potential function Ψ(Y,D), as follows: Ψ(Y,D...

    work page

  55. [55]

    Case1: UEi∈ Nchanges its computation offloading strategy while keeps the inference steps unchanged

    work page

  56. [56]

    Case2: UEi∈ Nchanges its inference steps while keeps the computation offloading strategy unchanged

    work page

  57. [57]

    Case1: Suppose that an UEiupdates its decisiony i toy ′ i (i.e.,y ik = 1→y ik′ = 1), that is, UEiswitches from an ESkto a new onek ′

    Case3: UEi∈ Nchanges both the computation offload- ing strategy and the inference step strategy. Case1: Suppose that an UEiupdates its decisiony i toy ′ i (i.e.,y ik = 1→y ik′ = 1), that is, UEiswitches from an ESkto a new onek ′. LetYandY ′ = (yi′,Y i)denote the strategy profiles before and after the strategy update of UEi, respectively. It is easy to se...

    work page

  58. [58]

    All the stable stationary points of the ODE in (40) are NE points ofG S1

    work page

  59. [59]

    Proof:Please refer to Theorem 3.2 in [48]

    All the NE points ofG S1 are stable stationary points of the ODE in (40). Proof:Please refer to Theorem 3.2 in [48]

    work page