Learning-Based Matching Game for Task Scheduling and Resource Collaboration in Intent-Driven Task-Oriented Networks

Abstract

With the rapid advancement of satellite communication technologies, space information networks (SINs) have become essential infrastructure for complex service delivery and cross-domain task coordination, facilitating the transition toward an intent-driven task-oriented coordination paradigm across the space, ground, and user segments. This study presents a novel intent-driven task-oriented network (IDTN) framework to address task scheduling and resource allocation challenges in SINs. The scheduling problem is formulated as a three-sided matching game that incorporates the preference attributes of entities across all network segments. To manage the variability of random task arrivals and dynamic resources, a context–aware linear upper-confidence-bound online learning mechanism is integrated to reduce decision-making uncertainty. Simulation results demonstrate the effectiveness of the proposed IDTN framework. Compared with conventional baseline methods, the framework achieves significant performance improvements, including a 4.4%–28.9% increase in average system reward, a 6.2%–34.5% improvement in resource utilization, and a 5.6%–35.7% enhancement in user satisfaction. The proposed framework is expected to facilitate the integration and orchestration of space-based platforms.