An ADMM Based Method for Computation Rate Maximization in Wireless Powered Mobile-Edge Computing Networks

Suzhi Bi, Ying-Jun Angela ZhangarXiv 2018 · arXiv

TL;DR

This paper presents a highly efficient optimization method based on the Alternating Direction Method of Multipliers (ADMM) to maximize the computation rate of wireless-powered IoT networks. By jointly optimizing energy transfer, time allocation, and binary offloading decisions, the proposed algorithm achieves near-optimal performance with linear computational complexity, making it ideal for large-scale deployments.

What Problem Does This Paper Address?

Internet of Things (IoT) devices are often hamstrung by two major factors: finite battery life and limited processing power. Two emerging technologies aim to solve this: Wireless Power Transfer (WPT), which charges devices via radio frequency signals, and Mobile-Edge Computing (MEC), which allows devices to offload heavy computations to a nearby server (like a 5G base station).

The challenge lies in integrating them. In a "wireless powered MEC" network, the Access Point (AP) must decide how to split a limited time frame between broadcasting energy and receiving offloaded data. Furthermore, the authors assume a binary offloading policy: a task is either computed locally on the device or offloaded entirely to the server. This creates a "combinatorial explosion"—with $N$ devices, there are $2^N$ possible offloading configurations. Finding the optimal balance between charging, local computing, and offloading is a non-convex, NP-hard problem that typically doesn't scale well as the number of devices increases.

Methods

The authors propose a centralized control scheme where the AP manages the network resources. The core of their approach is a reformulation of the optimization problem into a form solvable by ADMM (Alternating Direction Method of Multipliers).

  1. System Model: The network uses a Time-Division Duplexing (TDD) frame. The first part of the frame is dedicated to WPT (charging), and the second part is for task offloading (data transmission).
  2. Decomposition: The complex, coupled global problem is decomposed into $N$ parallel subproblems—one for each wireless device (WD). In each ADMM iteration, each WD independently finds its optimal local variables (computing mode and time).
  3. Coordination: The AP then updates "coupling variables" (system-wide time allocation) using a low-complexity bi-section search.
  4. Complexity: Unlike exhaustive search methods that grow exponentially ($2^N$), each iteration of this ADMM approach is $O(N)$. Crucially, the authors demonstrate that the number of iterations required for convergence remains nearly constant as the network grows, resulting in an overall linear complexity.
System model showing the AP integrated with an MEC server and RF transmitter, serving multiple WDs in different modes.

Datasets

The paper relies on numerical simulations rather than real-world datasets. The simulations use standard physical parameters:

  • Hardware: Powercast TX91501-3W transmitter (3W RF power) and P2110 Powerharvester (51% efficiency).
  • Channel Model: Free-space path loss model at 915 MHz.
  • Setup: Networks ranging from 10 to 30 devices, with distances and weights (priorities) assigned randomly or uniformly to test various scenarios.

Key Insights and Ideas

  • Binary vs. Partial Offloading: While many studies simplify the problem by allowing "partial offloading" (treating the decision as a continuous variable), this paper tackles the more realistic, harder binary case.
  • The Distance Threshold: A key finding is that the optimal strategy is highly dependent on channel quality. Devices close to the AP (strong channels) should almost always offload to leverage the server's power. Devices far away (weak channels) are better off computing locally because the energy cost and time required for data transmission outweigh the benefits of offloading.
  • Scalability for Massive IoT: Because the subproblems are solved in parallel and the complexity is linear, this method is uniquely suited for the "Massive IoT" vision of future 5G/6G networks.

Key Results

  • Near-Optimality: In small-scale tests where an exhaustive search was possible, the ADMM method achieved performance within 0.5% of the true mathematical optimum.
  • Superiority over Benchmarks: The proposed method significantly outperformed "offloading-only" and "local-computing-only" strategies. As shown in Figure 3, the ADMM approach adapts to changes in path loss and distance where fixed strategies fail.
  • Linear Complexity: Figure 5 confirms that the average number of iterations stays roughly constant (around 40-50) even as the number of devices triples, proving the $O(N)$ complexity claim.
Performance comparison showing ADMM tracking the optimal solution across different path loss exponents.

Limitations and Future Work

  • Channel Reciprocity: The model assumes the uplink and downlink channels are identical. While the authors note the algorithm can be extended to non-reciprocal cases, the current derivation relies on this simplification.
  • Overhead Neglect: The study neglects the time and energy spent on channel estimation and the time the AP takes to send results back to the devices. While often negligible in MEC, these could become bottlenecks in specific high-bandwidth or high-latency applications.
  • Static Scenarios: The optimization happens within a single time frame. Future work could explore dynamic scenarios where device mobility and fluctuating energy harvesting rates are considered over multiple frames.

Summary updated

Abstract

In this paper, we consider a wireless powered mobile edge computing (MEC) network, where the distributed energy-harvesting wireless devices (WDs) are powered by means of radio frequency (RF) wireless power transfer (WPT). In particular, the WDs follow a binary computation offloading policy, i.e., data set of a computing task has to be executed as a whole either locally or remotely at the MEC server via task offloading. We are interested in maximizing the (weighted) sum computation rate of all the WDs in the network by jointly optimizing the individual computing mode selection (i.e., local computing or offloading) and the system transmission time allocation (on WPT and task offloading). The major difficulty lies in the combinatorial nature of multi-user computing mode selection and its strong coupling with transmission time allocation. To tackle this problem, we propose a joint optimization method based on the ADMM (alternating direction method of multipliers) decomposition technique. Simulation results show that the proposed method can efficiently achieve near-optimal performance under various network setups, and significantly outperform the other representative benchmark methods considered. Besides, using both theoretical analysis and numerical study, we show that the proposed method enjoys low computational complexity against the increase of networks size.
BibTeX
@misc{bi2018admm,
  author = {Suzhi Bi and Ying-Jun Angela Zhang},
  title = {{An ADMM Based Method for Computation Rate Maximization in Wireless Powered Mobile-Edge Computing Networks}},
  year = {2018},
  eprint = {1802.04142},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1802.04142},
}

Loading related content...

Figure