Pith. sign in

REVIEW 5 major objections 5 minor 60 references

A Dynamic Service Offloading Algorithm Based on Lyapunov Optimization in Edge Computing

T0 review · 5 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that LDSO, an online drift-plus-penalty offloading algorithm, solves the long-term cost minimization in Eq. (1) with an O(1/V) cost bound, strong queue stability, and roughly 10% lower cost and 18.75% more offloaded data th

desk verdict Familiar Lyapunov offloading with a plausible greedy matching extension; the empirical story is reasonable, but the central O(ε) proof runs an inequality backwards and the stability theorem is nonstandard—major revision needed. read the letter →

arxiv 2509.10475 v1 pith:VNASR4VO submitted 2025-08-27 cs.NI

Peiyan Yuan , Ming Li , Chenyang Wang , Ledong An , Xiaoyan Zhao , Junna Zhang , Xiangyang Li , Huadong Ma This is my paper · ORCID
classification cs.NI
keywords edgecomputingserviceoffloadingdrift-plus-penaltyoptimizationqueuestabilitymulti-hopcollaborationcostminimizationgreedymatching
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tackles a horizontal edge-computing problem: many small base stations cooperate to serve user requests, but each server has a finite buffer, and the right split of work across servers is unclear when requests, channel rates, and processing rates change over time. It models the system as queues plus a per-slot cost that mixes energy and delay, then proposes LDSO, an online drift-plus-penalty offloading rule that needs no knowledge of future arrivals. The central claim is that LDSO drives the long-run average cost to within O(1/V) of the optimal value while keeping the queues strongly stable, with an O(1/ε²) convergence time when V=1/ε. Experiments on a real trace show about 10% lower cost and 18.75% more offloaded data than two recent baselines, so the paper intends to establish that stability-aware greedy matching is both provable and practical.

What carries the argument

The machinery is the Lyapunov function cluster: a quadratic Lyapunov function L(t)=1/2 Σ Q_i(t)², its one-slot drift ΔL(t), and the drift-plus-penalty ΔL(t)+V·Cost(t). Lemma 2 bounds this by B+V·Cost(t)+Σ Q_i(t) A_i(t), which turns the long-term stochastic problem into a per-slot deterministic minimization. The per-service cost C_i^k(t) is the object that carries the greedy matching: LDSO repeatedly selects the edge-server/service pair with minimal C_i^k(t), enforces Q_i(t)+A_i^k(t)≤Qmax_i, and updates queues via Q_i(t+1)=max{Q_i(t)−μ_i(t),0}+A_i(t).

What would settle it

Run LDSO on a workload whose arrival rate grows over time (for instance, doubling halfway through the experiment) while keeping the control parameter fixed; if the average cost exceeds C*+B/V or queues grow without bound, the stationarity premise is violated. A second check is to compare LDSO against an exact offline optimal policy on a small M×K instance; if the cost gap is much larger than O(1/V), the greedy matching is losing the optimality that the theorem assumes.

Watch

Extended reading notes

Core claim

The paper's central claim is that the long-term offloading cost minimization in Eq. (1), subject to queue stability and capacity constraints, can be solved online by minimizing, at each slot, a drift-plus-penalty expression. This yields a per-service cost C_i^k(t) = V θ E_i^k(t) + V(1−θ) T_i^k(t) + Q_i^k(t) A_i^k(t), and LDSO greedily picks the smallest such cost while respecting queue limits. Theorem 2 states that with ε=1/V and T≥1/ε² the time-averaged cost is within O(ε) of the optimal C*, and the arrival rate is within O(ε) of the processing rate. Theorem 3 bounds the average cost by C*+B/V, and Theorem 4 gives a strong-stability bound on the squared queue lengths. Simulated comparisons

Load-bearing premise

The proofs assume the random workload seen by each server is statistically the same from slot to slot, so a fixed long-run optimal policy exists; the paper relies on this through the drift-plus-penalty theorem without stating it, and if it fails the convergence and cost bounds do not automatically hold.

Editorial extensions

If this is right

  • An operator can tune V to trade cost against buffer occupancy: raising V pushes average cost down toward C* while allowing more data to sit in queues, and Theorem 4 keeps the time-averaged squared backlog bounded.
  • No statistical knowledge of arrivals is needed at runtime; LDSO only observes current queues, current request counts, and channel state, so it can run in changing conditions without retraining.
  • The O(1/ε²) convergence time means that after roughly 1/ε² slots the algorithm is within O(ε) of the optimal cost, which is useful for deciding how long to warm up.
  • The per-slot greedy matching is cheap, making the approach applicable to systems with many services and many edge servers.
  • If the queue cap is removed, cost improves further, so the model cleanly separates the price of stability from the cost of storage limits.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the proofs rely on drift-plus-penalty theory's stationarity assumption; under a non-stationary workload with trends in arrival rate, the O(ε) cost guarantee would need re-derivation, and a sliding-window or regret-based variant would be a natural test.
  • Editorial inference: the same drift-plus-penalty core works for any per-slot separable objective—replacing energy/delay terms with latency percentiles, carbon cost, or revenue would only change the terms inside C_i^k(t).
  • Editorial check: because the greedy matching is a heuristic, the paper's theoretical bound assumes the per-slot minimization is exact; comparing LDSO with an exact small-instance solver would separate the drift-plus-penalty benefit from greedy loss.
  • Editorial caution: the 18.75% offload improvement is measured on one urban trace; re-running on multiple traces with different request distributions would indicate how general the gain is.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper studies cost minimization in a horizontally collaborative edge-computing system. It formulates a long-term stochastic optimization problem (Eq. 1) with buffer-stability constraints, constructs an energy-plus-delay cost model (Eq. 2 and Section 4.2.2), and proposes LDSO, an online Lyapunov-optimization-based algorithm with a greedy matching heuristic (Algorithm 1). The paper claims an O(ε)-approximation with O(1/ε²) convergence (Theorem 2), an average-cost bound C*+B/V (Theorem 3), strong stability of the buffer queues (Theorem 4), and empirical improvements of about 10% lower cost and 18.75% more offloaded data over DSARA and MECNC (Section 6).

Significance. If the theoretical claims were correct, LDSO would be a useful contribution: it would show that a standard drift-plus-penalty construction can be adapted to multi-hop horizontal offloading with a tunable cost/stability trade-off, and the greedy matching would be an implementable per-slot solver. The paper also ships a reasonable level of model detail: multi-hop transmission, per-server queues, energy and delay components, and a Shanghai Telecom trace study. Credit goes to the authors for making the model and algorithm concrete. However, the central theoretical guarantees are not established. The missing stationarity premise, the mismatch between the aggregate and per-service drift terms, the reversed inequality in the proof of Theorem 2, and the non-standard and unproved stability statement in Theorem 4 are load-bearing. Absent these proofs, the paper is an empirical study of a heuristic with a plausible Lyapunov-inspired design, not a validated solution to Eq. (1).

major comments (5)
  1. [Section 4.1, Theorems 2-3] The drift-plus-penalty machinery of [13] requires the controlled stochastic primitives A_i(t), μ_i(t), r(t) to be stationary (typically i.i.d. over slots or at least ergodic) so that a stationary randomized policy with average cost C* exists. This assumption is never stated. Instead, Section 4.1 uses a time-varying service-request probability model P_k(t) from [48], and the simulations use non-stationary scenarios. Without stationarity or ergodicity, C* in Theorems 2 and 3 is not well-defined, and Eq. (29) has no target to converge to. This is a load-bearing gap, not a technicality.
  2. [Section 5.1, Lemma 2 and Theorem 1] The upper bound analyzed in Lemma 2, Eq. (26), contains the aggregate term Σ_i Q_i(t) A_i(t), where Q_i(t) and A_i(t) are the total queue length and total arrival at server i. Theorem 1, Eq. (27), replaces this with Σ_k Q_i^k(t) A_i^k(t), and Algorithm 1 optimizes the per-service expression. In general (Σ_k Q_i^k)(Σ_k A_i^k) ≠ Σ_k Q_i^k A_i^k, so the per-slot problem actually solved is not the minimization of the drift-plus-penalty upper bound used in the proofs. The optimality and stability bounds in Theorems 2 and 3 therefore do not apply to the implemented objective.
  3. [Theorem 2 proof, Eqs. (38)-(49)] Eq. (38) states C* ≤ \bar C(t) + Σ_i δ_i E{Q_i(T)}/T, a lower bound on \bar C(t). Eq. (49) then concludes \bar C(t) ≤ C* + O(ε), which is the reverse inequality and does not follow. The intermediate chain in Eqs. (40)-(43) bounds E{L(T)} in terms of ∥E{Q(T)}∥ and Eq. (45) bounds ∥E{Q(T)}∥, but none of these steps yields an upper bound on the average cost \bar C(t). Thus the O(ε) approximation and the claimed O(1/ε²) convergence time of Theorem 2 are unsupported, even if a stationarity assumption were added.
  4. [Section 5.3, Theorem 4] Strong stability of a queue is standardly defined as limsup_{T→∞} (1/T) Σ_{t=0}^{T-1} E{Q_i(t)} < ∞, as used in constraint (1a). Eq. (54) instead asserts limsup (1/T²) Σ_t Σ_i ½ Q_i(t)² < ∞, which is a different and weaker statement. Moreover, the proof is algebraically invalid: a bound on the endpoint term E{L(T)}−E{L(0)} cannot be converted to Eq. (55)'s bound on the time average of L(t) without additional assumptions; Eq. (55) also contains an unexplained factor T on the right-hand side. Theorem 4 does not establish Eq. (1a).
  5. [Section 5.2, Algorithm 1 and Theorems 2-3] The proofs of Theorems 2 and 3 presuppose that the per-slot decision minimizes the drift-plus-penalty expression. Algorithm 1 is a greedy matching heuristic (lines 12-19), and no lemma shows that it computes the minimizer of {C_i^k(t)} or even a constant-factor approximation to it. Without such a result, the theoretical trade-off is not tied to the implemented LDSO, and the simulations cannot be regarded as verifying Eq. (29) or Eq. (50) for the actual algorithm.
minor comments (5)
  1. [Eq. (1c)] Constraint (1c) is printed as KΣ n_i^k(t)b_k ≤ (Qmax_i − Q_i(t)) ≤ μ_i(t). The second inequality appears structurally odd and should be re-derived and restated.
  2. [Eq. (13)] The denominator 'µi(t)− ∑ nk_i(t) u=1 λk_ubk' requires parentheses; as written, if the arrival sum exceeds μ_i(t), the expression becomes negative and M/M/1 stability is not enforced.
  3. [Section 6, Figures 6-7] Figure 7(a,b) captions say 'System cost' while the text describes buffered data volume and the y-axis reads 'Data Backlogs'; the labels are inconsistent and should be corrected.
  4. [Section 6.2] The claims 'reduced cost is about an average of 10%' and '18.75 percent more offloaded data' are reported as point values without error bars, confidence intervals, or multiple simulation runs. Statistical support is missing.
  5. [Eq. (3)] A_i(t) depends on P_k(t), but the time-varying request-probability model from [48] is not defined in this manuscript. The authors should either define P_k(t) or provide a self-contained reference to the relevant equations.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity; the Lyapunov optimality proofs are imported from Neely and do not reduce to their inputs. A minor self-citation supplies the traffic model but is not load-bearing.

  1. other [Section 4.1, Eq. (3); Section 6.1]
    "This study uses a time-varying service request probability model [48] to calculate the probability of the k-th service type, rather than the traditional Zipf distribution."

    The arrival process A_i(t) is generated via P_k(t) taken from the authors' own prior work [48], and the simulations evaluate LDSO under a Poisson/time-varying request process of the same kind, giving the empirical validation a mild self-referential flavor. However, this self-citation is not load-bearing for the central optimality claims: Theorems 2-4 are derived from Neely's drift-plus-penalty framework and the supporting-hyperplane theorem, not from the specific traffic model. The self-citation merely supplies an input traffic process, so the cost/stability bounds do not reduce to it by construction.

full rationale

The paper's derivation chain starts from the cost minimization problem (1), constructs the Lyapunov drift in Lemma 1 from the queue update (5), and forms the drift-plus-penalty bound in Lemma 2. The greedy algorithm minimizes the per-slot cost C_i^k(t), and Theorems 2-4 are intended to follow from the standard Lyapunov optimization framework of Neely [13,49] and the supporting-hyperplane theorem [50]. These are external, independent mathematical tools, not self-citations or fitted parameters. The only notable self-citation is the time-varying request model [48] used to define A_i(t); it colors the simulation protocols but does not drive the theoretical guarantees. There are genuine proof concerns -- e.g., Theorem 2's proof uses Eq. (38) in the wrong direction, and Theorem 1 replaces the aggregate penalty sum_i Q_i(t)A_i(t) with per-service terms sum_i sum_k Q_i^k(t)A_i^k(t) -- but these are mathematical correctness issues, not circularity. No fitted constant is renamed as a prediction, and the experimental comparison is against external baselines DSARA and MECNC. Thus the central claims retain independent content; the self-citation is minor and not load-bearing.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

Axioms are mostly standard queueing/Lyapunov background plus domain assumptions about the service model. The stationarity assumption and the self-cited request model [48] are load-bearing. No invented physical entities are introduced.

free parameters (2)
  • θ (cost weight) = 0.5 in experiments
    Chosen, not optimized; controls the energy/delay trade-off in Cost(t) = θE_total + (1-θ)T_total (Eq. 2).
  • V (Lyapunov control parameter) = varied 200 to 12000 in Section 6
    Trade-off knob in drift-plus-penalty; not fitted, but the claimed O(1/V) cost and O(V) backlog depend on it.
assumptions (6)
  • domain assumption Arrival, service, and channel processes are stationary or i.i.d. across time slots
    Required for Neely's drift-plus-penalty optimality (cited [13]); never stated. Sections 5.1-5.2 use it implicitly in Theorems 2-3.
  • standard math Standard Lyapunov drift-plus-penalty optimality results from [13]
    The paper builds its cost bounds on Neely's framework (Lemmas 1-2, Theorem 3).
  • standard math Supporting hyperplane theorem [50] and Lagrange multiplier duality
    Invoked in the Theorem 2 proof, Eqs. (31)-(33).
  • domain assumption M/M/1 queueing formula for computation delay
    Equation (13) models each SBS's computation delay as 1/(μ_i(t) - Σλ), assuming Poisson arrivals; no validation is given.
  • domain assumption Time-varying service request probability model from [48] is valid
    n_i^k(t) = n_i × P_k(t) uses a self-cited model [48]; the distribution is not specified in this paper, and the simulations rely on it to generate arrivals.
  • domain assumption Terminal computing is negligible
    Explicitly stated in Section 4.1: 'it is reasonable to assume that IoT terminals have little computing capability, so terminal computing is not considered.'

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Dynamic Service Offloading Algorithm Based on Lyapunov Optimization in Edge Computing." pith.science (2026). https://pith.science/paper/VNASR4VO

@misc{pith2026250910475,
  author       = {Pith},
  title        = {Pith review of: A Dynamic Service Offloading Algorithm Based on Lyapunov Optimization in Edge Computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VNASR4VO}},
  note         = {Machine review of arXiv:2509.10475}
}
read the original abstract

This study investigates the trade-off between system stability and offloading cost in collaborative edge computing. While collaborative offloading among multiple edge servers enhances resource utilization, existing methods often overlook the role of queue stability in overall system performance. To address this, a multi-hop data transmission model is developed, along with a cost model that captures both energy consumption and delay. A time-varying queue model is then introduced to maintain system stability. Based on Lyapunov optimization, a dynamic offloading algorithm (LDSO) is proposed to minimize offloading cost while ensuring long-term stability. Theoretical analysis and experimental results verify that the proposed LDSO achieves significant improvements in both cost efficiency and system stability compared to the state-of-the-art.

Figures

Figures reproduced from arXiv: 2509.10475 by the authors.

Figure 1
Figure 1. Edge service offloading architecture. base stations (SBS), and n users. M SBSs are covered by the BS. Un￾like most existing single-hop service offloading, tasks can be finished by the cooperative edge server, and multiple SBSs jointly provide services to users through horizontal cooperation. When users send a service request to a nearby SBS, if the SBS deploys the user’s re￾quested service, it processes the service … view at source ↗
Figure 2
Figure 2. Edge server buffer queue model. where µi(t)represents the processing rate of the i-th edge server, and Equation (5) states that the service data volume offloaded in time slot t will be stored in the queue and processed in the next time slot. If the amount of data processing in a time slot is less than the amount of data received. Let µ max i represent the maximum workload of the edge server. At each time slot, the a… view at source ↗
Figure 3
Figure 3. LDSO algorithm framework The Lyapunov drift function at time slot t is then changed as follows. ∆L(t) =E (XM i=1 1 2 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: System stability analysis for different V and λ. 6.2.2 The influence of the control parameter V on the system cost 0 2 4 6 8 10 Control Parameter V 105 0 2 4 6 8 10 12 System Cost LDSO DSARA MECNC (a) System cost w/ Qmax i = 4000 0 2 4 6 8 10 Control Parameter V 105 0 …
Figure 6
Figure 6. Figure 6: Performance comparison of LDSO, DSARA, and MECNC under different control parameters. We now compare LDSO with the state-of-the-art works. The com￾parison algorithms are DSARA [41] and MECNC [44]. Firstly, we analyze the system cost of the three algorithms [PITH_FULL_I…
Figure 4
Figure 4. Figure 4: Deployment of edge servers This study uses the Shanghai Telecom dataset to simulate an edge service offloading system as shown in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 7
Figure 7. Figure 7: Performance comparison of LDSO, DSARA, and MECNC under different control parameters over time. with larger V , consistent with Theorem 4 (Equation 54), and a higher V allows more data to be cached. Among the three, LDSO consis￾tently maintains the smallest buffer size,…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages

  1. [48]

    Maximizing the Capacity of Edge- Caching Networks With User-Content Evolution Relationshi p

    Y uan, Peiyan and Li, Shuhong and Cai, Y unyun and Zhao, Xiaoyan and Tang, Shaojie and Li, Xiangyang. Maximizing the Capacity of Edge- Caching Networks With User-Content Evolution Relationshi p. IEEE Transactions on V ehicular Technology, 71(11), 12169–12178, 2022

    work page 2022

  2. [13]

    Stochastic network optimization wit h application to communication and queueing systems

    Neely, Michael J. Stochastic network optimization wit h application to communication and queueing systems. Synthesis Lectures on Commu- nication Networks, 3(1), 1–211, 2010

    work page 2010

  3. [1]

    Treasure collection on foggy islands: Building secure netw ork archives for Internet of Things

    Duan, Huayi and Zheng, Yifeng and Wang, Cong and Y uan, Xin gliang. Treasure collection on foggy islands: Building secure netw ork archives for Internet of Things. IEEE Internet of Things Journal, 6(2), 2637–2650, 2018

    work page 2018

  4. [2]

    Accuracy Rate Maximization in Edge Federate d Learn- ing With Delay and Energy Constraints

    Y uan, Peiyan and Huang, Rong and Zhang, Junna and Zhang, E n and Zhao, Xiaoyan. Accuracy Rate Maximization in Edge Federate d Learn- ing With Delay and Energy Constraints. IEEE Systems Journal , 17(2), 2053–2064, 2023

    work page

  5. [3]

    Lyapunov-based partial computatio n offloading for multiple mobile devices enabled by harvested energy in M EC

    Guo, Min and Wang, Wei and Huang, Xing and Chen, Y anru and Z hang, Lei and Chen, Liangyin. Lyapunov-based partial computatio n offloading for multiple mobile devices enabled by harvested energy in M EC. IEEE Internet of Things Journal, 9(11), 9025–9035, 2021

    work page 2021

  6. [4]

    An online algorithm for task offloading in h eteroge- neous mobile clouds

    Zhou, Bowen and Dastjerdi, Amir V ahid and Calheiros, Rod rigo N and Buyya, Rajkumar. An online algorithm for task offloading in h eteroge- neous mobile clouds. ACM Transactions on Internet Technology (TOIT), 18(2), 1–25, 2018

    work page 2018

  7. [5]

    A taxonomy and survey of energy-efficient da ta cen- ters and cloud computing systems

    Beloglazov, Anton and Buyya, Rajkumar and Lee, Y oung Cho on and Zomaya, Albert. A taxonomy and survey of energy-efficient da ta cen- ters and cloud computing systems. Advances in computers, 82 , 47–111, 2011

    work page 2011

  8. [6]

    Collaboration improves the capacity of m obile edge computing

    Y uan, Peiyan and Cai, Y unyun and Huang, Xiaoyan and Tang, Shaojie and Zhao, Xiaoyan. Collaboration improves the capacity of m obile edge computing. IEEE Internet of Things Journal, 6(6), 10610–10 619, 2019

    work page 2019

Show all 60 references
  1. [7]

    Exploiting smart e-Health gateways at the edge of heal thcare Internet-of-Things: A fog computing approach

    Rahmani, Amir M and Gia, Tuan Nguyen and Negash, Behailu a nd An- zanpour, Arman and Azimi, Iman and Jiang, Mingzhe and Liljeb erg, Pasi. Exploiting smart e-Health gateways at the edge of heal thcare Internet-of-Things: A fog computing approach. Future Gene ration Com- puter S...

    2018

  2. [8]

    A mobility-aware cross-edge computa- tion offloading framework for partitionable applications

    Zhao, Hailiang and Deng, Shuiguang and Zhang, Cheng and D u, Wei and He, Qiang and Yin, Jianwei. A mobility-aware cross-edge computa- tion offloading framework for partitionable applications. In 2019 IEEE International Conference on W eb Services (ICWS), pp. 193–200, 2019

    2019

  3. [9]

    IoTScal-H: hybrid monitoring solution based on cloud compu ting for au- tonomic middleware-level scalability management within I oT systems and different SLA traffic requirements

    Zyane, Abdellah and Bahiri, Mohamed Nabil and Ghammaz, A bdelilah. IoTScal-H: hybrid monitoring solution based on cloud compu ting for au- tonomic middleware-level scalability management within I oT systems and different SLA traffic requirements. International Jour nal of Commu...

    2020

  4. [10]

    Cloud-vision: Real-time fa ce recog- nition using a mobile-cloudlet-cloud acceleration archit ecture

    Soyata, Tolga and Muraleedharan, Rajani and Funai, Col in and Kwon, Minseok and Heinzelman, Wendi. Cloud-vision: Real-time fa ce recog- nition using a mobile-cloudlet-cloud acceleration archit ecture. In 2012 IEEE symposium on computers and communications (ISCC) , pp. 000059–0...

    2012

  5. [11]

    Caching hit ratio maximization in mobile edge computing wit h node co- operation

    Y uan, Peiyan and Shao, Saike and Geng, Lijuan and Zhao, X iaoyan. Caching hit ratio maximization in mobile edge computing wit h node co- operation. Computer Networks, 200, 108507, 2021

    2021

  6. [12]

    Optimal energy efficiency with delay con- straints for multi-layer cooperative fog computing networks

    Vu, Thai T and Nguyen, Diep N and Hoang, Dinh Thai and Dutk iewicz, Eryk and Nguyen, Thuy V . Optimal energy efficiency with delay con- straints for multi-layer cooperative fog computing networks. IEEE Trans- actions on Communications, 69(6), 3911–3929, 2021

    2021

  7. [14]

    Joint task assignment and resource allocation for D2 D-enabled mobile-edge computing

    Xing, Hong and Liu, Liang and Xu, Jie and Nallanathan, Ar u- mugam. Joint task assignment and resource allocation for D2 D-enabled mobile-edge computing. IEEE Transactions on Communicatio ns, 67(6), 4193–4207, 2019

    2019

  8. [15]

    Dynamic task offloading and resource allocation f or ultra- reliable low-latency edge computing

    Liu, Chen-Feng and Bennis, Mehdi and Debbah, Merouane a nd Poor, H Vincent. Dynamic task offloading and resource allocation f or ultra- reliable low-latency edge computing. IEEE Transactions on Communi- cations, 67(6), 4132–4150, 2019

    2019

  9. [16]

    Joint task offloading a nd resource al- location for multi-server mobile-edge computing networks

    Tran, Tuyen X and Pompili, Dario. Joint task offloading a nd resource al- location for multi-server mobile-edge computing networks . IEEE Trans- actions on V ehicular Technology, 68(1), 856–868, 2018

    2018

  10. [17]

    Dynamic c om- putation offloading for mobile-edge computing with energy h arvesting devices

    Mao, Y uyi and Zhang, Jun and Letaief, Khaled B. Dynamic c om- putation offloading for mobile-edge computing with energy h arvesting devices. IEEE Journal on Selected Areas in Communications, 34(12), 3590–3605, 2016

    2016

  11. [18]

    Computation offloading scheduling for periodic tasks in mobile edge computing

    Jošilo, Sla ¯dana and Dán, György. Computation offloading scheduling for periodic tasks in mobile edge computing. IEEE/ACM Trans actions on Networking, 28(2), 667–680, 2020

    2020

  12. [19]

    Energy efficient dynamic offloading in mobile edge computing for internet of things

    Chen, Ying and Zhang, Ning and Zhang, Y ongchao and Chen, Xin and Wu, Wen and Shen, Xuemin. Energy efficient dynamic offloading in mobile edge computing for internet of things. IEEE Transact ions on Cloud Computing, 9(3), 1050–1060, 2019

    2019

  13. [20]

    eTime: Energy-efficient trans mission between cloud and mobile devices

    Shu, Peng and Liu, Fangming and Jin, Hai and Chen, Min and Wen, Feng and Qu, Y upeng and Li, Bo. eTime: Energy-efficient trans mission between cloud and mobile devices. In 2013 Proceedings IEEE INFO- COM, pp. 195–199, 2013

    2013

  14. [21]

    A survey on delay-aware resource control f or wire- less systems¡ªLarge deviation theory, stochastic Lyapuno v drift, and dis- tributed stochastic learning

    Cui, Ying and Lau, Vincent KN and Wang, Rui and Huang, Hua ng and Zhang, Shunqing. A survey on delay-aware resource control f or wire- less systems¡ªLarge deviation theory, stochastic Lyapuno v drift, and dis- tributed stochastic learning. IEEE Transactions on Inform ation The...

    2012

  15. [22]

    EMM: Energy-awa re mo- bility management for mobile edge computing in ultra dense n etworks

    Sun, Y uxuan and Zhou, Sheng and Xu, Jie. EMM: Energy-awa re mo- bility management for mobile edge computing in ultra dense n etworks. IEEE Journal on Selected Areas in Communications, 35(11), 2637–2646, 2017

    2017

  16. [23]

    Col- laborative content placement among wireless edge caching s tations with time-to-live cache

    Chen, Lixing and Song, Linqi and Chakareski, Jacob and X u, Jie. Col- laborative content placement among wireless edge caching s tations with time-to-live cache. IEEE transactions on multimedia, 22(2 ), 432–444, 2019

    2019

  17. [24]

    Efficiency of Buffer Caching in Computi ng-Intensive Workloads

    Bahn, Hyokyung. Efficiency of Buffer Caching in Computi ng-Intensive Workloads. In 2020 7th International Conference on Information Science and Control Engineering (ICISCE), pp. 548–552, 2020

    2020

  18. [25]

    Accuracy-guaranteed collaborative DNN infe rence in in- dustrial IoT via deep reinforcement learning

    Wu, Wen and Y ang, Peng and Zhang, Weiting and Zhou, Congh ao and Shen, Xuemin. Accuracy-guaranteed collaborative DNN infe rence in in- dustrial IoT via deep reinforcement learning. IEEE Transac tions on In- dustrial Informatics, 17(7), 4988–4998, 2020

    2020

  19. [26]

    Distribute d opti- mization for computation offloading in edge computing

    Lin, Rongping and Zhou, Zhijie and Luo, Shan and Xiao, Y o ng and Wang, Xiong and Wang, Sheng and Zukerman, Moshe. Distribute d opti- mization for computation offloading in edge computing. IEEE Transac- tions on Wireless Communications, 19(12), 8179–8194, 2020

    2020

  20. [27]

    Joint task offloading and data caching in mobile edge computing net works

    Zhang, Ni and Guo, Songtao and Dong, Yifan and Liu, Defan g. Joint task offloading and data caching in mobile edge computing net works. Computer Networks, 182, 107446, 2020

    2020

  21. [28]

    Fog and IoT: An overview of r esearch opportunities

    Chiang, Mung and Zhang, Tao. Fog and IoT: An overview of r esearch opportunities. IEEE Internet of things journal, 3(6), 854– 864, 2016

    2016

  22. [29]

    Efficient algorithms for capacitated cloudle t place- ments

    Xu, Zichuan and Liang, Weifa and Xu, Wenzheng and Jia, Mi ke and Guo, Song. Efficient algorithms for capacitated cloudle t place- ments. IEEE Transactions on Parallel and Distributed Syste ms, 27(10), 2866–2880, 2015

    2015

  23. [30]

    A survey on computation offloading modeli ng for edge computing

    Lin, Hai and Zeadally, Sherali and Chen, Zhihong and Lab iod, Houda and Wang, Lusheng. A survey on computation offloading modeli ng for edge computing. Journal of Network and Computer Applicatio ns, 169, 102781, 2020

    2020

  24. [31]

    A mobility-aware vehicular caching scheme in content centric networks: Model and optimization

    Zhang, Y ao and Li, Changle and Luan, Tom Hao and Fu, Y uchu an and Shi, Weisong and Zhu, Lina. A mobility-aware vehicular caching scheme in content centric networks: Model and optimization. IEEE T ransactions on V ehicular Technology, 68(4), 3100–3112, 2019

    2019

  25. [32]

    EEDTO: an energy-efficient dynamic task offloading algorithm for blockchain-enabled IoT-edge-cloud orches- trated computing

    Wu, Huaming and Wolter, Katinka and Jiao, Pengfei and De ng, Yingjun and Zhao, Y ubin and Xu, Minxian. EEDTO: an energy-efficient dynamic task offloading algorithm for blockchain-enabled IoT-edge-cloud orches- trated computing. IEEE Internet of Things Journal, 8(4), 21 63–2176, 2020

    2020

  26. [33]

    Follow me at the ed ge: Mobility-aware dynamic service placement for mobile edge c omputing

    Ouyang, Tao and Zhou, Zhi and Chen, Xu. Follow me at the ed ge: Mobility-aware dynamic service placement for mobile edge c omputing. IEEE Journal on Selected Areas in Communications, 36(10), 2333–2345, 2018

    2018

  27. [34]

    Caching assisted correlated ta sk offloading for IoT devices in mobile edge computing

    Tang, Chaogang and Zhu, Chunsheng and Wu, Huaming and Li u, Chun- yan and Rodrigues, Joel JPC. Caching assisted correlated ta sk offloading for IoT devices in mobile edge computing. In 2021 IEEE Global Com- munications Conference (GLOBECOM), pp. 1–6, 2021

    2021

  28. [35]

    Learning-aided computation of floading for trusted collaborative mobile edge computing

    Li, Y uqing and Wang, Xiong and Gan, Xiaoying and Jin, Hai ming and Fu, Luoyi and Wang, Xinbing. Learning-aided computation of floading for trusted collaborative mobile edge computing. IEEE Tran sactions on Mobile Computing, 19(12), 2833–2849, 2019

    2019

  29. [36]

    A task offloading algorithm forcloud- edge collaborative system based on Lyapunov optimization

    Gao, Jixun and Chang, Rui and Y ang, Zhipeng and Huang, Qu anzheng and Zhao, Y uanyuan and Wu, Y u. A task offloading algorithm forcloud- edge collaborative system based on Lyapunov optimization. Cluster Computing, 26(1), 337–348, 2023

    2023

  30. [37]

    Online collaborative dat a caching in edge computing

    Xia, Xiaoyu and Chen, Feifei and He, Qiang and Grundy, Jo hn and Abdelrazek, Mohamed and Jin, Hai. Online collaborative dat a caching in edge computing. IEEE Transactions on Parallel and Distributed Systems, 32(2), 281–294, 2020

    2020

  31. [38]

    Dynamic Servic e Mi- gration and Request Routing for Microservice in Multicell M obile-Edge Computing

    Chen, Xiangyi and Bi, Y uanguo and Chen, Xueping and Zhao , Hai and Cheng, Nan and Li, Fuliang and Cheng, Wenlin. Dynamic Servic e Mi- gration and Request Routing for Microservice in Multicell M obile-Edge Computing. IEEE Internet of Things Journal, 9(15), 13126–13143, 2022

    2022

  32. [39]

    Research o n offload- ing strategy in edge computing of internet of things

    Duan, Xiaoting and Xu, Fei and Sun, Y ongyong. Research o n offload- ing strategy in edge computing of internet of things. In 2020 Inter- national Conference on Computer Network, Electronic and Au tomation (ICCNEA), pp. 206–210, 2020

    2020

  33. [40]

    Distributed and dynamic service placement in perv asive edge computing networks

    Ning, Zhaolong and Dong, Peiran and Wang, Xiaojie and Wa ng, Shu- peng and Hu, Xiping and Guo, Song and Qiu, Tie and Hu, Bin and Kwok, Ricky YK. Distributed and dynamic service placement in perv asive edge computing networks. IEEE Transactions on Parallel and Distributed Sys- ...

    2020

  34. [41]

    Service characteristics-oriented joi nt optimization of radio and computing resource allocation in mobile-edge c omputing

    Feng, Jie and Liu, Lei and Pei, Qingqi and Hou, Fen and Y an g, Tingt- ing and Wu, Jinsong. Service characteristics-oriented joi nt optimization of radio and computing resource allocation in mobile-edge c omputing. IEEE Internet of Things Journal, 8(11), 9407–9421, 2021

    2021

  35. [42]

    Multi-user Computation Offlo ading Al- gorithm for Mobile Edge Computing

    Pan, Meini and Li, Zhihua. Multi-user Computation Offlo ading Al- gorithm for Mobile Edge Computing. In 2021 2nd International Con- ference on Electronics, Communications and Information Te chnology (CECIT), pp. 771–776, 2021

    2021

  36. [43]

    Mobility-aware offloading and resource allo cation in MEC-enabled IoT networks

    Hu, Han and Song, Weiwei and Wang, Qun and Zhou, Fuhui and Hu, Rose Qingyang. Mobility-aware offloading and resource allo cation in MEC-enabled IoT networks. In 2020 16th International Conference on Mobility, Sensing and Networking (MSN) , pp. 554–560, 2020

    2020

  37. [44]

    Mobile edge computing network control: Tradeoff be tween de- lay and cost

    Cai, Y ang and Llorca, Jaime and Tulino, Antonia M and Mol isch, An- dreas F. Mobile edge computing network control: Tradeoff be tween de- lay and cost. In GLOBECOM 2020-2020 IEEE Global Communications Conference, pp. 1–6, 2020

    2020

  38. [45]

    Joint service caching and c ompu- tation offloading to maximize system profits in mobile edge-c loud com- puting

    Fan, Qingyang and Lin, Junyu and Feng, Guangsheng and Ga o, Zihan and Wang, Huiqiang and Li, Y afei. Joint service caching and c ompu- tation offloading to maximize system profits in mobile edge-c loud com- puting. In 2020 16th international conference on mobility, sensing an d ...

    2020

  39. [46]

    Power-delay tradeoff in multi-user mobile-edge computing systems

    Mao, Y uyi and Zhang, Jun and Song, SH and Letaief, Khaled Ben. Power-delay tradeoff in multi-user mobile-edge computing systems. In 2016 IEEE global communications conference (GLOBECOM) , pp. 1–6, 2016

    2016

  40. [47]

    Energy efficiency and delay tradeoff in an me c- enabled mobile iot network

    Hu, Han and Song, Weiwei and Wang, Qun and Hu, Rose Qingya ng and Zhu, Hongbo. Energy efficiency and delay tradeoff in an me c- enabled mobile iot network. IEEE Internet of Things Journal , 9(17), 15942–15956, 2022

    2022

  41. [49]

    A simple convergence time analysis of drift-plus- penalty for stochastic optimization and convex programs

    Neely, Michael J. A simple convergence time analysis of drift-plus- penalty for stochastic optimization and convex programs. a rXiv preprint arXiv:1412.0791, 2014

    2014 arXiv

  42. [50]

    Convex analysis and optimization

    Bertsekas, Dimitri and Nedic, Angelia and Ozdaglar, As uman. Convex analysis and optimization. Athena Scientific, 2003

    2003

  43. [51]

    Dependency-A ware Microservice Deployment for Edge Computing: A Deep Reinfor cement Learning Approach with Network Representation

    Wang, Chenyang and Y u, Hao and Li, Xiuhua and Ma, Fei and W ang, Xiaofei and Taleb, Tarik and Leung, Victor CM. Dependency-A ware Microservice Deployment for Edge Computing: A Deep Reinfor cement Learning Approach with Network Representation. IEEE Trans actions on Mobile Comp...

    2024

  44. [52]

    Heterogene ous Edge Caching Based on Actor-Critic Learning With Attention Mecha- nism Aiding

    Wang, Chenyang and Li, Ruibin and Wang, Xiaofei and Tale b, Tarik and Guo, Song and Sun, Y uxia and Leung, Victor CM. Heterogene ous Edge Caching Based on Actor-Critic Learning With Attention Mecha- nism Aiding. IEEE Transactions on Network Science and Engin eering, 10(6), 340...

    2023

  45. [53]

    Dynamic task allocation and service migration in edge-clou d iot system based on deep reinforcement learning

    Chen, Y an and Sun, Y anjing and Wang, Chenyang and Taleb, Tarik. Dynamic task allocation and service migration in edge-clou d iot system based on deep reinforcement learning. IEEE Internet of Thin gs Journal, 9(18), 16742–16757, 2022

    2022

  46. [54]

    Multitask offloading strate gy opti- mization based on directed acyclic graphs for edge computin g

    Chen, Jiawen and Y ang, Y ajun and Wang, Chenyang and Zhan g, Heng and Qiu, Chao and Wang, Xiaofei. Multitask offloading strate gy opti- mization based on directed acyclic graphs for edge computin g. IEEE Internet of Things Journal, 9(12), 9367–9378, 2021

    2021

  47. [55]

    Hierarchical Deep Reinforcement Learning for Joint Service Caching and Computation Offloadi ng in Mo- bile Edge-Cloud Computing

    Sun, Chuan and Li, Xiuhua and Wang, Chenyang and He, Qian g and Wang, Xiaofei and Leung, Victor CM. Hierarchical Deep Reinforcement Learning for Joint Service Caching and Computation Offloadi ng in Mo- bile Edge-Cloud Computing. IEEE Transactions on Services Computing, 2024

    2024

  48. [56]

    J oint Task and Computing Resource Allocation in Distributed Edge Comp uting Systems via Multi-Agent Deep Reinforcement Learning

    Chen, Y an and Sun, Y anjing and Y u, Hao and Taleb, Tarik. J oint Task and Computing Resource Allocation in Distributed Edge Comp uting Systems via Multi-Agent Deep Reinforcement Learning. IEEE Trans- actions on Network Science and Engineering, 2024

    2024

  49. [57]

    Mission-Aware UA V Deployment for Post-Disaster Scen arios: A Worst-Case SAC-Based Approach

    Wang, Jingjing and Sun, Y anjing and Wang, Bowen and Ushio, Toshim- itsu. Mission-Aware UA V Deployment for Post-Disaster Scen arios: A Worst-Case SAC-Based Approach. IEEE Transactions on V ehic ular Technology, 2023

    2023

  50. [58]

    Determinist ic latency/jitter-aware service function chaining over beyo nd 5G edge fab- ric

    Y u, Hao and Taleb, Tarik and Zhang, Jiawei. Determinist ic latency/jitter-aware service function chaining over beyo nd 5G edge fab- ric. IEEE Transactions on Network and Service Management, 1 9(3), 2148–2162, 2022

    2022

  51. [59]

    Wang, Chenyang and Y u, Hao and Li, Xiuhua and Ma, Fei and W ang, Xiaofei and Taleb, Tarik and Leung, Victor C. M.. Dependency - Aware Microservice Deployment for Edge Computing: A Deep Re in- forcement Learning Approach With Network Representation. IEEE Transactions on Mobile...

    2024

  52. [60]

    Joint UA V Deployment and Edg e Association for Energy-Efficient Federated Learning

    Wu, Tao and Li, Maomao and Qu, Y uben and Wang, Hongjun and Wei, Zhenhua and Cao, Jiannong. Joint UA V Deployment and Edg e Association for Energy-Efficient Federated Learning. IEEE Transac- tions on Cognitive Communications and Networking, 1–1, 202 5. DOI: 10.1109/TCCN.2025.3543365

    2025

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.