Abstract: As a member of unmanned intelligent marine vehicles, autonomous underwater vehicles（AUVs）play a significant role in applying network science to complex tasks such as AUV navigation and obstacle avoidance to achieve automated execution. This paper proposes an end-to-end AUV obstacle avoidance algorithm based on event-triggered reinforcement learning, in which an environmental perception model is designed to assess the relative position relationship between the AUV and either single or multiple unknown static obstacles and target points. Additionally, two different event-trigger mechanisms are integrated when customizing the state space and reward function. Finally, obstacle avoidance experiments are conducted on a simulation platform to validate its effectiveness.