Low-Altitude Satellite-AAV Collaborative Joint Mobile Edge Computing and Data Collection via Diffusion-based Deep Reinforcement Learning

TL;DR

This work addresses the challenge of jointly optimizing MEC and DC in a satellite–AAV network under dynamic channels and limited resources. It introduces QAGOB, a diffusion-model–enabled DRL framework that recasts the problem as an MDP with a transformed action space and leverages GS-based GD association, QVPO-based diffusion policy, and entropy regularization. The approach achieves higher MEC task completion rates, greater DC data collection, and lower AAV energy consumption than five benchmarks, with stable convergence across varying scenarios. The results demonstrate the practical potential of integrated satellite–AAV MEC-DC systems for remote or disaster environments, enabling rapid deployment and efficient resource usage.

Abstract

The integration of satellite and autonomous aerial vehicle (AAV) communications has become essential for the scenarios requiring both wide coverage and rapid deployment, particularly in remote or disaster-stricken areas where the terrestrial infrastructure is unavailable. Furthermore, emerging applications increasingly demand simultaneous mobile edge computing (MEC) and data collection (DC) capabilities within the same aerial network. However, jointly optimizing these operations in heterogeneous satellite-AAV systems presents significant challenges due to limited on-board resources and competing demands under dynamic channel conditions. In this work, we investigate a satellite-AAV-enabled joint MEC-DC system where these platforms collaborate to serve ground devices (GDs). Specifically, we formulate a joint optimization problem to minimize the average MEC end-to-end delay and AAV energy consumption while maximizing the collected data. Since the formulated optimization problem is a non-convex mixed-integer nonlinear programming (MINLP) problem, we propose a Q-weighted variational policy optimization-based joint AAV movement control, GD association, offloading decision, and bandwidth allocation (QAGOB) approach. Specifically, we reformulate the optimization problem as an action space-transformed Markov decision process to adapt the variable action dimensions and hybrid action space. Subsequently, QAGOB leverages the multi-modal generation capacities of diffusion models to optimize policies and can achieve better sample efficiency while controlling the diffusion costs during training. Simulation results show that QAGOB outperforms five other benchmarks, including traditional DRL and diffusion-based DRL algorithms. Furthermore, the MEC-DC joint optimization achieves significant advantages when compared to the separate optimization of MEC and DC.

Figures (8)

• Figure 1: Illustration of the LEO satellite-AAV-enabled joint MEC-DC system. The system enables multi-tier computing where tasks are processed by AAVs or further offloaded to the satellite. The satellite also acts as the destination for data collection. Note that inter-cell interference exists among AAVs during the simultaneous service provisioning.
• Figure 2: The schematic of the proposed QAGOB approach. The architecture integrates a diffusion model into the DRL actor-critic structure to serve as the policy network, whereby the diffusion policy generates complex continuous actions via a reverse denoising process. Furthermore, the device association is decoupled from the neural network and efficiently solved by a GS-based approach.
• Figure 3: Comparison results of QAGOB approach and benchmarks. (a) Average E2E delay. (b) Collected data volume. (c) AAV energy consumption.
• Figure 4: Comparison results of MEC task completion rate and DC rate of QAGOB approach and benchmarks.
• Figure 5: Reward curves of QAGOB approach and benchmarks.