Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cost-Efficient Computation Offloading in SAGIN: A Deep Reinforcement Learning and Perception-Aided Approach

Yulan Gao, Ziqiang Ye, Han Yu

TL;DR

This work tackles cost-efficient computation offloading in Space-Air-Ground Integrated Networks (SAGIN) under mobility and sensing uncertainty. It introduces a DRL-and-Perception-aided online approach that fuses Lyapunov drift-plus-penalty optimization with deep reinforcement learning (DDPG for continuous UAV decisions and DQN for UAV-BS associations) and a perception module (mmWave radar plus YOLOv7 vision) to enhance state estimation. The framework decomposes the problem into three subproblems—task hosting/offloading, association control, and BS resource allocation—solved with DRL and SGHS, incorporating ${\mathcal F}( {\boldsymbol H}(t) ) = \Delta{\mathcal L}({\boldsymbol H}(t)) + V {\mathbb E}{\{ \mathcal G(t) | {\boldsymbol H}(t) \}}$ to balance cost and stability. Simulations show the proposed approach achieves lower time-averaged network cost and better stability than baselines across dynamic SAGIN scenarios, underscoring its practical potential for 6G systems with perception-enabled edge orchestration.

Abstract

The Space-Air-Ground Integrated Network (SAGIN), crucial to the advancement of sixth-generation (6G) technology, plays a key role in ensuring universal connectivity, particularly by addressing the communication needs of remote areas lacking cellular network infrastructure. This paper delves into the role of unmanned aerial vehicles (UAVs) within SAGIN, where they act as a control layer owing to their adaptable deployment capabilities and their intermediary role. Equipped with millimeter-wave (mmWave) radar and vision sensors, these UAVs are capable of acquiring multi-source data, which helps to diminish uncertainty and enhance the accuracy of decision-making. Concurrently, UAVs collect tasks requiring computing resources from their coverage areas, originating from a variety of mobile devices moving at different speeds. These tasks are then allocated to ground base stations (BSs), low-earth-orbit (LEO) satellite, and local processing units to improve processing efficiency. Amidst this framework, our study concentrates on devising dynamic strategies for facilitating task hosting between mobile devices and UAVs, offloading computations, managing associations between UAVs and BSs, and allocating computing resources. The objective is to minimize the time-averaged network cost, considering the uncertainty of device locations, speeds, and even types. To tackle these complexities, we propose a deep reinforcement learning and perception-aided online approach (DRL-and-Perception-aided Approach) for this joint optimization in SAGIN, tailored for an environment filled with uncertainties. The effectiveness of our proposed approach is validated through extensive numerical simulations, which quantify its performance relative to various network parameters.

Cost-Efficient Computation Offloading in SAGIN: A Deep Reinforcement Learning and Perception-Aided Approach

TL;DR

This work tackles cost-efficient computation offloading in Space-Air-Ground Integrated Networks (SAGIN) under mobility and sensing uncertainty. It introduces a DRL-and-Perception-aided online approach that fuses Lyapunov drift-plus-penalty optimization with deep reinforcement learning (DDPG for continuous UAV decisions and DQN for UAV-BS associations) and a perception module (mmWave radar plus YOLOv7 vision) to enhance state estimation. The framework decomposes the problem into three subproblems—task hosting/offloading, association control, and BS resource allocation—solved with DRL and SGHS, incorporating to balance cost and stability. Simulations show the proposed approach achieves lower time-averaged network cost and better stability than baselines across dynamic SAGIN scenarios, underscoring its practical potential for 6G systems with perception-enabled edge orchestration.

Abstract

The Space-Air-Ground Integrated Network (SAGIN), crucial to the advancement of sixth-generation (6G) technology, plays a key role in ensuring universal connectivity, particularly by addressing the communication needs of remote areas lacking cellular network infrastructure. This paper delves into the role of unmanned aerial vehicles (UAVs) within SAGIN, where they act as a control layer owing to their adaptable deployment capabilities and their intermediary role. Equipped with millimeter-wave (mmWave) radar and vision sensors, these UAVs are capable of acquiring multi-source data, which helps to diminish uncertainty and enhance the accuracy of decision-making. Concurrently, UAVs collect tasks requiring computing resources from their coverage areas, originating from a variety of mobile devices moving at different speeds. These tasks are then allocated to ground base stations (BSs), low-earth-orbit (LEO) satellite, and local processing units to improve processing efficiency. Amidst this framework, our study concentrates on devising dynamic strategies for facilitating task hosting between mobile devices and UAVs, offloading computations, managing associations between UAVs and BSs, and allocating computing resources. The objective is to minimize the time-averaged network cost, considering the uncertainty of device locations, speeds, and even types. To tackle these complexities, we propose a deep reinforcement learning and perception-aided online approach (DRL-and-Perception-aided Approach) for this joint optimization in SAGIN, tailored for an environment filled with uncertainties. The effectiveness of our proposed approach is validated through extensive numerical simulations, which quantify its performance relative to various network parameters.
Paper Structure (39 sections, 1 theorem, 39 equations, 6 figures, 1 table, 2 algorithms)

This paper contains 39 sections, 1 theorem, 39 equations, 6 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Computation Offloading and Resource Allocation in SAGIN
  4. Perception and Communication Integration
  5. System Model
  6. Network Settings
  7. Task Model
  8. Communication Model
  9. Communication Rate Between Air and Ground
  10. Communication Rate Between Air and Satellite
  11. Dynamic State Perception Model
  12. Models for Computation Offloading and Associated Cost Analysis
  13. Computation Offloading Model
  14. Cost Model
  15. Cost of Computation Offloading Model ①
  16. ...and 24 more sections

Key Result

Theorem 1

Given that $V\geq 0$ and considering the network state in time slot $t$, represented by ${\boldsymbol H}(t)$, then where $\Pi$ is a positive constant.

Figures (6)

  • Figure 1: Architecture of SAGIN and the entire process of deep reinforcement learning and perception-aided approach for SAGIN.
  • Figure 2: Effectiveness of the proposed algorithms for subproblems.
  • Figure 3: Convergence of the overall optimization process combining three subproblems ( P1, P2, P3), i.e., Algorithm \ref{['algorithm:2']}.
  • Figure 4: Performance of time-averaged network cost.
  • Figure 5: Performance of time-averaged queue backlogs at UAVs and BSs.
  • ...and 1 more figures

Theorems & Definitions (1)

  • Theorem 1