Sep 2026· IEEE Internet of Things Journal· Vol 13, pp. 40830-40845· 0 citations· 37 references
Abstract
Mobile edge computing (MEC) has emerged as a promising paradigm to support latency-sensitive applications by deploying computing and storage resources closer to end users. Service caching at the network edge is an effective approach to further reduce response latency and improve quality of service (QoS). However, the increasing scale and dynamic nature of user requests pose significant challenges to efficient service caching and updating under limited edge resources and cost budgets. In this article, we investigate the joint optimization problem of service caching and dynamic updating in a cloud–edge–end collaborative architecture, with the objective of minimizing the average service access latency of users. We first construct a comprehensive system model that captures communication latency, service response delay, and the costs associated with service caching and updates. The problem is formulated as a nonlinear integer programming problem and addressed via a two-stage solution framework. Specifically, in the initial caching stage, we employ Lagrangian relaxation to reduce the high-dimensional knapsack problem and design an efficient algorithm integrated with Tabu Search. In the dynamic updating stage, we develop a Monte Carlo tree search (MCTS)-based algorithm enhanced by service popularity awareness, along with a novel service replacement strategy that considers user preference and service reliability. Extensive experiments under various system scales demonstrate that the proposed approach significantly outperforms state-of-the-art methods in terms of average service latency, cost budget satisfaction, and update efficiency.
An adaptive Beta-policy and delayed-update multi-agent soft actor-critic method, abbreviated as ABDMASAC, which uses a Beta policy to model bounded actions and achieves a better overall trade-off than the selected MASAC-backbone and on-policy MARL baselines under the considered simulation settings.
Zheng Yao, Jie Liu, Changjun Deng et al.· Computers, Materials & C...· 0 citations
We consider an edge caching system with a finite capacity edge-cache connected to a backend server via a wireless channel. The backend server stores the latest versions of dynamic contents. Users request the edge server for the contents, which can either fetch fresh versions from backend and serve or can serve locally cached versions or can even deny service. The edge server must decide which items to cache due to limited capacity. Fetching from the backend server incurs a fetching cost, serving a stale version incurs an ageing cost proportional to the content’s age-of-version (AoV), and denying service incurs a missing cost. We address optimal content fetching, caching and delivery problem to minimize the expected time average cost. The optimal control problem, a Markov decision process (MDP), suffers from curse-of-dimensionality. We frame the problem as a restless multiarmed bandit (RMAB), show that it is indexable, and design a Whittle index based joint policy for content fetching, caching and delivery. We provide explicit expressions for the Whittle indices. Finally, we demonstrate that our proposed policy performs very close to optimal.
Ankita Koley, Chandramani Singh· ACM Transactions on Modeling...· 0 citations
Satellite-Terrestrial Integrated Networks (STINs) leverage the global reach of satellite systems to push onboard computing and caching resources toward the network edge, enabling truly ubiquitous, anytime-anywhere services for remote Internet of Things (IoT) applications. In this context, efficiently orchestrating the constrained onboard computing and caching resources under stochastic service demands in a scalable, low-complexity, and resilient manner remains a critical challenge. Existing solutions primarily rely on centralized optimization or multi-agent learning techniques, which struggle to cope with the non-stationarity arising from the interdependence of autonomous agent decisions. In this work, we take a step further and develop a distributed game-theoretic framework for joint task offloading and service caching in STINs, providing provable equilibrium guarantees. Specifically, the joint problem is formulated as a non-cooperative stochastic game among IoT devices that autonomously determine their computing, association, and satellite caching strategies to minimize their end-to-end latency subject to energy and cache capacity constraints. The formulated game is proven to converge to a Correlated Equilibrium (CE), which generalizes the Nash Equilibrium (NE) to correlated, probabilistic strategy profiles across devices. Two distributed no-regret learning algorithms, operating under different information availability and rationality regimes, are introduced to derive the CE. The effectiveness and efficiency of the two no-regret learning algorithms are validated through extensive simulations, considering alternative equilibria, learning-based methods, baseline computing schemes, and varying network and algorithm configurations.
Filothei Linardatou, Maria Diamanti, E. Tsiropoulou et al.· IEEE Open Journal of the Com...· 0 citations
Simulation results confirm that the proposed JORC framework substantially reduces latency, energy consumption, and overall system cost, while increasing the successful task completion ratio compared to existing baseline approaches.
Tanmay Baidya, S. Moh· Italian National Conference...· 0 citations
To address the problems of limited caching resources at edge nodes, dynamically changing user requests, and imbalanced node loads in mobile edge computing, this paper proposes a hedonic-game-based cooperative caching optimization strategy (HG-CC). First, a joint utility function integrating caching benefit and cooperation cost is constructed by considering content popularity, node resource status, network distance, cache complementarity, and cooperation cost. Second, the caching cooperation relationship among edge nodes is modeled as a hedonic game, enabling nodes to autonomously select cooperative coalitions according to their own utilities and form a stable coalition partition. Finally, a differentiated content deployment strategy is adopted within each coalition to reduce redundant caching and improve content coverage and cache resource utilization. Simulation and real-world request trace experiments demonstrate that, compared with the local most popular caching strategy and the fixed-neighbor cooperative caching strategy, the proposed method improves the cache hit ratio while reducing the average service latency and node load variance. It also exhibits good robustness against content popularity prediction errors, validating its effectiveness and adaptability.
Ting Wang, Jun Zhang, Ruoheng Chen· 2026 8th International Confe...· 0 citations
Proactive caching in multi-tier cellular networks (MTCNs) is an efficient technique for alleviating heavy data traffic on backhaul links, thereby reducing content latency and increasing the overall cache hit ratio (CHR) and user satisfaction ratio (USR). The CHR and USR can be further enhanced by incorporating cache-enabled device-to-device (D2D) and cloud cache into MTCNs, forming cloud and D2D-assisted multi-tier cellular networks (CD2DMTCNs). In this context, we formulate two problems: the first is the joint optimization of the overall CHR and USR of the system, while the second focuses on minimizing the average content latency. The popularity prediction of newly published content using machine learning algorithms has become increasingly attractive, with support vector machines (SVMs) showing promising results in this regard. However, for improved performance, an uncertainty assessment is necessary. To address this, bootstrapping is applied to SVM to obtain confidence and prediction intervals. Additionally, K-means clustering is used to form virtual clusters within both D2D networks and MTCNs, facilitating content popularity prediction. This enables the maximization of CHR and USR through a clustered-bootstrapped support vector machine (clustered-BSVM). In addition, establishing criteria for content delivery is essential. Since multiple cache-enabled transmitters may store the content requested by users, only one transmitter can deliver the requested content at a time. This constraint is addressed using the branch-and-bound method of linear programming, which minimizes the average latency of the system. Numerical simulations demonstrate that proactive caching based on clustered-BSVM outperforms existing methods in terms of CHR and USR, while the branch-and-bound method effectively resolves the content delivery problem, minimizing system latency.
Ayaz Ahmad, Fawad Ahmad, Muhammad Suleman Khan et al.· PeerJ Computer Science· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduSep 2, 2026
What if pathology foundation models could do more with less? GigaPath-Flash and GigaTIME-Flash cut computational demands while maintaining strong performance, opening the door to larger studies and broader exploration. The post GigaPath-Flash and GigaTIME-Flash: Toward population-scale discovery with efficient pathology foundation models appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduAug 31, 2026
With millions of users across the world, Julia has been used to conduct cutting-edge research and to design new drugs, jet engines, heat pumps, and more.
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
