Continuum robots, due to their characteristics of flexible and large deformations, as well as dexterous mobility, have become a development trend for enhancing safety and interactivity of robots in the future. Digital twin is an important technical support to achieve coexistence and integration among robots, environments and humans. This article focuses on the research of tensegrity flexible manipulator and combines technologies such as digital twin and virtual simulation to enable deep integration of the tensegrity manipulator in virtual and physical spaces. A data communication architecture is established to achieve realtime data transmission, thereby enhancing the collaborative efficiency between the flexible manipulator and humans. Moreover, dynamic obstacle avoidance in complex environments is considered through collision detection feedback. Furthermore, a tensegrity flexible manipulator digital twin system based on dynamics is developed, and the effectiveness of the constructed system is verified by the experiments on bidirectional control of the tensegrity manipulator between virtual and physical domains. This research provides a reference for remote intelligent monitoring and control of other robots.