Abstract: 6G Autonomous Networks needs to realize the network automation full scene demand-aware service for multi-layer users. Operators' users urgently need a method to effectively mine the multi-layer user intents and realize the automatic on-demand allocation of resources. Therefore, a fully automated framework for managing network resources by turning user intention into strategy is proposed. Firstly, considering the scarcity of intent mining data, a method of using unlabeled corpus to improve the ability of intent entity mining is proposed. Secondly, comprehensively considering the network service quality and user business needs, the deep reinforcement learning algorithm is used to optimize and manage the division of network resources, improve the user experience, balance the network load and maximize the utilization of resources. Experimental results show that the proposed framework can mine users' intents more rapidly and divide network resources more accurately to ensure the quality of service.

Key words: the Autonomous Networks, intent mining, demand-aware service, deep reinforcement learning