Energy Efficient Resource Allocation in UAV-Enabled Mobile Edge Computing Networks

Zhaohui Yang, Zhaohui Yang 0001, Cunhua Pan, Kezhi Wang, Mohammad Shikh-BahaeiIEEE Trans. Wireless Communications 2019 · DOI

TL;DR

This paper tackles the challenge of minimizing energy consumption in UAV-enabled mobile edge computing (MEC) networks by jointly optimizing user association, power control, computation capacity, and UAV location. They propose an iterative algorithm that breaks down the complex problem into manageable subproblems, achieving significant power savings compared to existing methods.

What Problem Does This Paper Address?

The increasing demand for mobile data and computation-intensive applications has led to the exploration of Unmanned Aerial Vehicles (UAVs) as mobile edge computing (MEC) servers. UAVs offer advantages like flexible deployment and high connectivity, especially in areas lacking traditional infrastructure. However, efficiently managing resources in such dynamic environments is crucial to minimize energy consumption, which is a significant concern for battery-powered UAVs and user devices. This paper addresses the problem of minimizing the total power consumption (UEs and UAVs) in a multi-UAV MEC network by jointly optimizing several key aspects: which users connect to which UAV (user association), how much power UEs use to transmit data (power control), how much computation capacity each UAV allocates (computation capacity allocation), and where the UAVs should be positioned (location planning). The core challenge lies in the non-convexity and complexity of this joint optimization problem, especially with multiple UAVs and discrete decision variables.

Methods

The authors propose an iterative algorithm to solve the complex, non-convex sum power minimization problem. This algorithm iteratively optimizes four key components:

  • User Association: This is tackled using a compressive sensing-based algorithm that approximates the $\ell_0$-norm with a sequence of weighted $\ell_1$-norm minimizations. This approach helps in handling the discrete nature of user association.
  • Power Control: The optimal transmission power for UEs is derived based on the current user association, computation capacity, and UAV locations.
  • Computation Capacity Allocation: This subproblem is decomposed into smaller, manageable optimization problems. The optimal computation capacity for UAVs is derived in closed-form.
  • Location Planning: The optimal 3D location and beamwidth of the UAVs are found using a one-dimensional search method.

To ensure a feasible starting point for this iterative process, a fuzzy c-means (FCM) clustering-based algorithm is employed to obtain an initial feasible solution.

Datasets

Not discussed in the paper. The paper focuses on a simulation-based evaluation of the proposed algorithm.

Key Insights and Ideas

  • Joint Optimization is Key: The paper highlights that optimizing individual components (like just location or just association) is insufficient. Jointly optimizing user association, power control, computation capacity, and UAV location is crucial for significant energy savings.
  • Decomposition for Tractability: The complex problem is effectively broken down into smaller, solvable subproblems, making it computationally feasible. This iterative approach allows for finding a near-optimal solution.
  • Compressive Sensing for Association: Using compressive sensing techniques to handle the discrete user association problem is an innovative approach that simplifies the optimization.
  • FCM for Initialization: Employing FCM clustering to find an initial feasible solution is a practical strategy for dealing with the non-convexity and potential infeasibility of the problem.
  • UAVs as Flexible Resources: The work demonstrates how the flexibility of UAVs (location, altitude, beamwidth) can be leveraged to significantly improve network efficiency and reduce power consumption.

Key Results

The proposed iterative algorithm (IACL) demonstrates superior performance compared to benchmark methods like SCAEAH (fixed altitude/beamwidth) and ECC (equal computation capacity allocation).

  • Power Reduction: The algorithm achieves significant reductions in total network power consumption, with numerical results showing a reduction from over 1000 W to around 420 W after a few iterations.
  • Convergence Speed: The algorithm converges rapidly, often within three iterations, indicating its efficiency.
  • Performance Gains: IACL consistently outperforms SCAEAH and ECC across various scenarios, especially in low maximal computation capacity and high data rate conditions.
  • Near-Optimal Solution: The proposed IACL algorithm approaches the performance of an exhaustive search (EXH) method, which is computationally very expensive, suggesting it finds a near-globally optimal solution.
  • Impact of Parameters: The results show that increasing maximal latency leads to lower power consumption, and increasing computation capacity of UAVs also reduces overall power consumption by enabling more offloading.

Limitations and Future Work

  • Computational Complexity: While the proposed algorithm is efficient, solving the subproblems still involves iterative processes and optimization techniques, which can be computationally intensive for very large-scale networks.
  • Practical Deployment: The paper focuses on theoretical optimization and simulation. Real-world deployment challenges like dynamic channel conditions, UAV battery life management, and interference mitigation are not extensively covered.
  • Future Work: The authors suggest exploring the optimization problem for UAV-enabled MEC networks where UAVs themselves act as users, which is a novel direction for future research. They also mention that the optimization problem for UAV-enabled MEC network, where UAVs are served as UEs, is left for future work.

Summary updated

BibTeX
@article{yang2019energy,
  author = {Zhaohui Yang and Zhaohui Yang and Cunhua Pan and Kezhi Wang and Mohammad Shikh-Bahaei},
  title = {{Energy Efficient Resource Allocation in UAV-Enabled Mobile Edge Computing Networks}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2019},
  doi = {10.1109/twc.2019.2927313},
}
Figure