Joint Flexible Beamforming, Resource Allocation, and Task Buffering for Movable-Antenna Enabled MEC Systems
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.
Rosa: I'm Rosa.
Dev: And I'm Dev, here with Taro.
Rosa: We're looking at the paper: "Joint Flexible Beamforming, Resource Allocation, and Task Buffering for Movable-Antenna Enabled MEC Systems".
Dev: Here's the gist of the paper. The paper proposes a novel MEC framework empowered by movable antennas (MAs) and task buffering to address the limitations of conventional fixed-position antenna systems and rigid single-slot task execution models in mobile edge computing environments The system enables continuous antenna repositioning to maximize array gains and decouples task generation from immediate execution via buffer queues, allowing for flexible multi-slot scheduling System Architecture and Modeling The considered MEC system consists of one BS enhanced by MAs, equipped with an MEC server to provide computing services for the MUs, and each MU is equipped with a single FPA that is moving on the road The BS is enhanced by MAs, and it is also equipped with an MEC server to provide computing services for the MUs The set of MAs and MUs are denoted as M =
1, · · ·, M: and K =
1, · · ·, K: , respectively The two-dimensional (2D) location of the BS is
xBS, yBS: , and the height of the BS is H The position of the m-th MA at time slot t is given by xm(t) The channel coefficient between the k-th MU and the BS is given by hk→BS(t) The received signal at the BS is written as y(t) = X K k=1 √pkhk→BS(t)sk(t) + n(t), where n(t) =
n1(t), n2(t), · · ·, nM(t): T denotes the noise vector The offloading data rate of the k-th MU can be given by rk(t) = B log2 (1 + γk(t)), where B is the system bandwidth Optimization Framework and Decoupling The problem is formulated as minimizing the total energy consumption for data processing by jointly optimizing the receive beamforming w =
wk(t), ∀k, t: , the MA positions x =
xm(t), ∀m, t: , and the computing resource scheduling F =
fk(t), ∀k, t, F(t), ∀t: To transform the long-term stochastic problem into deterministic per-slot optimization subproblems, Lyapunov optimization is utilized This transformation involves constructing a Lyapunov function L(C(t)) and conditional Lyapunov drift ∇L(C(t)) to decouple the problem The resulting deterministic optimization problem (P3-1) is equivalent to a formulation where the objective function is minimized Optimization Subproblems and Algorithms The deterministic optimization problem (P3-1) is solved by decomposing it into three subproblems using an efficient alternating optimization (AO) algorithm 1.
Rosa: Receive Beamforming Optimization: This involves solving problem (P4-2), which is an SDR problem, using semi-definite relaxation (SDR) and successive convex approximation (SCA) to find the optimal receive beamforming vector Wk 2.
Dev: MA Position Optimization: This subproblem is solved using the projected gradient ascent (PGA) method to find a local optimal solution for the antenna positions x 3.
Rosa: Computing Resource Allocation: This problem, which is convex, can be solved by standard methods like the interior-point method to determine the optimal computing resource allocation F Performance and Stability Analysis The convergence of the proposed algorithm is guaranteed by Theorem 1, which shows that the long-term buffer stability of MUs and the MEC server holds Furthermore, Theorem 2 confirms that within each time slot t, the proposed AO algorithm converges to a stationary point of problem P3-1 The analysis also demonstrates that the average energy consumption is upper bounded The simulation results show that the MA-enabled scheme significantly outperforms the FPA-based scheme in terms of energy consumption and the sum computed data The proposed scheme always achieves the lowest average energy consumption Computational Complexity The overall computational complexity of Algorithm 1 over T time slots can be expressed as O T IAO KM3 + IPGAK2M2 + (K + 1)3 This complexity is determined by the three subproblems solved in each AO iteration, namely the receive beamforming optimization problem (P4-2), the MA position optimization problem (P5-1), and the computing resource allocation problem (P6) Tradeoffs and Control Parameters The balance coefficient Γ controls the tradeoff between energy consumption and task backlog, as validated in Section V Fig.
Title and authors: Rosa: The authors, Hao Wang, Chunjie Wang, Lingtong Ma, Wenchao Liu, Kan Yu, Jinke Ren, Yanyan Shen and Xuhui Zhang developed this framework to handle the energy-efficient and adaptable needs of these modern mobile edge computing systems.
Dev: They focus on using movable antennas or MAs to get better spatial flexibility while also decoupling when a task is generated from when it actually gets executed.
Taro: I'm wondering how continuous antenna repositioning actually helps with the uplink communication part, since most research in this area has been focused on downlink transmission scenarios.
Rosa: That's a fair point, Taro. The paper points out that existing research has mainly focused on downlink transmission scenarios, and they are specifically looking at the gap in uplink transmission for MEC architectures where users offload computation tasks to edge servers.
Dev: Exactly. They're trying to bridge that gap by fully utilizing the spatial degrees of freedom of MAs to realize flexible beamforming, which is what we need for user-to-server communication.
The paper's summary: Rosa: To summarize the core idea of "Joint Flexible Beamforming, Resource Allocation, and Task Buffering for Movable-Antenna Enabled MEC Systems," they propose a novel MEC framework that uses movable antennas and task buffering to solve the problems of limited spatial flexibility and rigid scheduling.
Dev: They are jointly optimizing three things: receive beamforming, the positions of the movable antennas themselves, and how computing resources are scheduled over time.
Taro: So it's not just about optimizing one part of the system; it’s a full joint optimization across all those variables simultaneously to minimize long-term energy consumption.
Rosa: Right. And they use Lyapunov optimization to turn this long-term stochastic problem into simpler, deterministic per-slot subproblems, which makes it tractable for real systems.
The paper's improvements: Dev: One major improvement they highlight is the decoupling of task generation from immediate execution through task buffering. This allows them to schedule computations over multiple time slots instead of being forced to finish everything in one slot.
Taro: That decoupling is key for handling those bursty workloads, like when a generative AI application suddenly generates a lot of tasks, giving the system the freedom to schedule things over the rest of the time interval.
Rosa: And they also focus on continuous antenna repositioning to maximize array gains and spatial diversity, which directly enhances uplink task offloading performance by optimizing receive beamforming vector of each user is aligned with the normalized instantaneous channel vector.
Dev: That adaptive transmission capability is what lets them maximize those antenna gains while maintaining link reliability in a dynamic environment.
Conclusion: Rosa: So, to wrap up "Joint Flexible Beamforming, Resource Allocation, and Task Buffering for Movable-Antenna Enabled MEC Systems," the main implication is that this framework offers a way to achieve better energy efficiency by jointly optimizing those three things—beamforming, antenna positions, and resource allocation.
Dev: It shows how task buffering can allow for flexible multi-slot scheduling which balances energy use and latency in a very dynamic mobile edge computing setup.
Taro: From my side, what I see is that they successfully managed the queue stability by using a Lyapunov function to bound the drift plus energy penalty, which means the system maintains strong stability even with unpredictable traffic.
Rosa: So it's about achieving this energy-queue tradeoff achieved by the proposed scheme while ensuring stability under unpredictable traffic conditions.
Dev: Yeah, it’s a solid approach for designing MEC systems that need to handle those highly dynamic workloads without sacrificing long-term energy efficiency or user experience.
Department of Networks, China Mobile Communications Group Co. · Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · University of Chinese Academy of Sciences
eess.SP, cs.IT, math.IT
Submitted: 2026-11-05
Updated: 2026-11-05
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 75/100
The gist: The paper proposes a novel MEC framework empowered by movable antennas (MAs) and task buffering to address the limitations of conventional fixed-position antenna systems and rigid single-slot task…
Key concepts
- Movable Antennas (MAs)
- These are antennas that can be physically moved on the road. The system uses them to continuously reposition themselves to maximize signal strength and diversity, which helps improve how well users can send data to the base station.
- Task Buffering
- This involves using queues at both the base station and mobile users. It decouples when a task is created from when it is actually processed, allowing the system to handle sudden bursts of data efficiently without immediate execution bottlenecks.
- Lyapunov Optimization
- This mathematical technique transforms a complex long-term optimization problem into simpler, manageable per-slot problems. It uses a specific function to track and bound the system's behavior over time, ensuring stability and guiding the system toward an optimal solution.
- Alternating Optimization (AO)
- This is an efficient algorithm used to solve the main optimization problem by breaking it down into three smaller, simpler subproblems: beamforming, antenna placement, and resource allocation. It solves these pieces sequentially until a good overall solution is found.
Terminology
Summary
The paper proposes a novel MEC framework empowered by movable antennas (MAs) and task buffering to address the limitations of conventional fixed-position antenna systems and rigid single-slot task execution models in mobile edge computing environments The system enables continuous antenna repositioning to maximize array gains and decouples task generation from immediate execution via buffer queues, allowing for flexible multi-slot scheduling
System Architecture and Modeling
The considered MEC system consists of one BS enhanced by MAs, equipped with an MEC server to provide computing services for the MUs, and each MU is equipped with a single FPA that is moving on the road The BS is enhanced by MAs, and it is also equipped with an MEC server to provide computing services for the MUs The set of MAs and MUs are denoted as M = [1, · · ·, M] and K = [1, · · ·, K], respectively The two-dimensional (2D) location of the BS is [xBS, yBS], and the height of the BS is H The position of the m-th MA at time slot t is given by xm(t) The channel coefficient between the k-th MU and the BS is given by hk→BS(t) The received signal at the BS is written as y(t) = X K k=1 √pkhk→BS(t)sk(t) + n(t), where n(t) = [n1(t), n2(t), · · ·, nM(t)]T denotes the noise vector The offloading data rate of the k-th MU can be given by rk(t) = B log2 (1 + γk(t)), where B is the system bandwidth
Optimization Framework and Decoupling
The problem is formulated as minimizing the total energy consumption for data processing by jointly optimizing the receive beamforming w = [wk(t), ∀k, t], the MA positions x = [xm(t), ∀m, t], and the computing resource scheduling F = [fk(t), ∀k, t, F(t), ∀t] To transform the long-term stochastic problem into deterministic per-slot optimization subproblems, Lyapunov optimization is utilized This transformation involves constructing a Lyapunov function L(C(t)) and conditional Lyapunov drift ∇L(C(t)) to decouple the problem The resulting deterministic optimization problem (P3-1) is equivalent to a formulation where the objective function is minimized
Optimization Subproblems and Algorithms
The deterministic optimization problem (P3-1) is solved by decomposing it into three subproblems using an efficient alternating optimization (AO) algorithm
-
Receive Beamforming Optimization: This involves solving problem (P4-2), which is an SDR problem, using semi-definite relaxation (SDR) and successive convex approximation (SCA) to find the optimal receive beamforming vector Wk
-
MA Position Optimization: This subproblem is solved using the projected gradient ascent (PGA) method to find a local optimal solution for the antenna positions x
-
Computing Resource Allocation: This problem, which is convex, can be solved by standard methods like the interior-point method to determine the optimal computing resource allocation F
Performance and Stability Analysis
The convergence of the proposed algorithm is guaranteed by Theorem 1, which shows that the long-term buffer stability of MUs and the MEC server holds Furthermore, Theorem 2 confirms that within each time slot t, the proposed AO algorithm converges to a stationary point of problem P3-1 The analysis also demonstrates that the average energy consumption is upper bounded The simulation results show that the MA-enabled scheme significantly outperforms the FPA-based scheme in terms of energy consumption and the sum computed data The proposed scheme always achieves the lowest average energy consumption
Computational Complexity
The overall computational complexity of Algorithm 1 over T time slots can be expressed as O T IAO KM3 + IPGAK2M2 + (K + 1)3 This complexity is determined by the three subproblems solved in each AO iteration, namely the receive beamforming optimization problem (P4-2), the MA position optimization problem (P5-1), and the computing resource allocation problem (P6)
Tradeoffs and Control Parameters
The balance coefficient Γ controls the tradeoff between energy consumption and task backlog, as validated in Section V Fig. 8 depicts the average energy consumption versus the time slot index under different balance coefficients Γ A larger balance coefficient assigns a higher weight to the energy minimization term in the drift-plus-penalty objective, driving the system to adopt more energy-efficient decisions Fig. p. 5
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Adaptive Spatial Beamforming for Enhanced Link Quality: The system can continuously reposition movable antennas to maximize array gains and spatial diversity, which directly enhances uplink task offloading performance by optimizing
receive beamforming vector of each user is aligned with the normalized instantaneous channel vector.
-
Flexible Multi-Slot Task Scheduling: The integration of task buffering allows the system to decouple
task generation from immediate execution via buffer queues,
enabling flexible multi-slot scheduling to handle bursty workloads like generative AI and low altitude economy services. -
Energy-Aware Resource Allocation: The framework minimizes long-term energy consumption by jointly optimizing receive beamforming, MA positions, and computing resources through a Lyapunov optimization approach that transforms the long-term problem into tractable per-slot subproblems.
-
Queue Stability Management: The use of task buffer queues at both the base station and mobile users ensures
strong stability
of these buffers, maintaining bounded average queue lengths under unpredictable traffic conditions by employing a Lyapunov function to bound the drift plus energy penalty. -
Dynamic Control Tradeoff: By adjusting the balance coefficient Gamma, the system can flexibly operate on a tradeoff curve between energy minimization and queue backlog, allowing design choices based on practical requirements such as balancing
energy-queue tradeoff achieved by the proposed scheme.
Related papers
- Open-H-RF: A Large-Scale Ultrasound Raw Channel Dataset Initiative
- Scalable Attention for 5G NR Channel Estimation
- Low-Altitude Wireless Networks: A Comprehensive Survey
- Sensing-Assisted OFDM with Index Modulation: A Waveform for Optical Wireless Integrated Communication and Localisation
- Iteration-Free Candidate Pruning and an Index-Domain VLSI Architecture for Sort-Free MIMO Box Decoding up to 1024-QAM
- A Foundation Model for Large-Scale Wireless Network Planning, Operation and Optimization