Digital Twin-Assisted Space-Air-Ground Integrated Multi-Access Edge Computing for Low-Altitude Economy: An Online Decentralized Optimization Approach

TL;DR

This work tackles the integration of space-air-ground MEC for the low-altitude economy by introducing a digital twin–assisted SAGIMEC framework and a QoS-focused joint optimization (JSC4OP). It develops an online decentralized optimization approach (ODOA) that combines Lyapunov problem transformation, online latency prediction via MAB/UCB in a digital twin, and a decentralized Stackelberg game for ISD and UAV decisions. Theoretical analysis confirms UAV energy constraint satisfaction and low per-slot complexity, while simulations show ODOA outperforms baselines in ISD cost, task latency, and energy metrics. The results demonstrate practical gains for LAE applications with scalable, real-time resource management and trajectory control in SAGIMEC networks.

Abstract

The emergence of space-air-ground integrated multi-access edge computing (SAGIMEC) networks opens a significant opportunity for the rapidly growing low altitude economy (LAE), facilitating the development of various applications by offering efficient communication and computing services. However, the heterogeneous nature of SAGIMEC networks, coupled with the stringent computational and communication requirements of diverse applications in the LAE, introduces considerable challenges in integrating SAGIMEC into the LAE. In this work, we first present a digital twin-assisted SAGIMEC paradigm for LAE, where digital twin enables reliable network monitoring and management, while SAGIMEC provides efficient computing offloading services for Internet of Things sensor devices (ISDs). Then, a joint satellite selection, computation offloading, communication resource allocation, computation resource allocation and UAV trajectory control optimization problem (JSC4OP) is formulated to maximize the quality of service (QoS) of ISDs. Given the complexity of JSC4OP, we propose an online decentralized optimization approach (ODOA) to address the problem. Specifically, JSC4OP is first transformed into a real-time decision-making optimization problem (RDOP) by leveraging Lyapunov optimization. Then, to solve the RDOP, we introduce an online learning-based latency prediction method to predict the uncertain system environment and a game theoretic decision-making method to make real-time decisions. Finally, theoretical analysis confirms the effectiveness of the ODOA, while the simulation results demonstrate that the proposed ODOA outperforms other alternative approaches in terms of overall system performance.

Figures (5)

• Figure 1: The proposed digital twin-assisted SAGIMEC architecture consists of a physical entity layer and a digital twin layer.
• Figure 2: The framework of ODOA.
• Figure 3: The impact of time slots on system performance. (a) Time-average ISD cost (Cost). (b) Average task completion latency (Latency). (c) Time-average ISD energy consumption (ISD-EC). (d) Time-average UAV energy consumption (UAV-EC).
• Figure 4: The impact of task data size on system performance. (a) Time-average ISD cost (Cost). (b) Average task completion latency (Latency). (c) Time-average ISD energy consumption (ISD-EC). (d) Time-average UAV energy consumption (UAV-EC).
• Figure 5: The impact of UAV computation resources on system performance. (a) Time-average ISD cost (Cost). (b) Average task completion latency (Latency). (c) Time-average ISD energy consumption (ISD-EC). (d) Time-average UAV energy consumption (UAV-EC).

Theorems & Definitions (22)

• Theorem 1
• proof
• Theorem 2
• proof
• Theorem 3
• proof
• Theorem 4
• proof
• Theorem 5
• proof