<p>Industry 5.0 places the combination of human decision-making and machine technology at the heart of high industrial automation. The real-time hybrid tasks resilient collaborative scheduling are the challenging problems as resilient and human-centric systems through human–machine collaboration. Current mainstream research usually perform manufacturing task scheduling and computing task scheduling separately to pursue their specific goals. Exist collaborative scheduling methods mainly focus on improving scheduling performance, ignore the dependencies between manufacturing and computing tasks, and the highly dynamic nature of the edge environment. In this paper, we propose a hybrid task scheduling model for the resilient collaborative scheduling problem of manufacturing and computing tasks. Firstly, based on the dependency between manufacturing tasks and computing tasks, we design a multi-agent graph reinforcement learning algorithm to balance the productivity and computational latency. Then, we propose an adaptive learning framework based on meta-learning for the highly dynamic edge computing environment, which fully utilizes the fast environment learning capability of meta-learning, enabling the algorithm to converge rapidly with fewer training steps when the environment changes. Finally, experiments show that our method can improve the performance efficiently.</p>