Soft robotics has recently become an emergent research area due to its unique characteristics compared to conventional rigid robots. Its inherent properties, such as compliance and flexibility, provide some promising characteristics for the current robotic applications, including safe human machine interaction, great adaptability to unknown environments and so on. On the contrary, its intrinsic 'soft' characteristic would result in some complex non-linear behaviors, causing more difficulties in deriving and deducing kinematic or dynamic models of soft robots than what we often do for rigid robots. In this paper, using Finite Element Method (FEM), we demonstrate a gain-scheduling closed-loop method to control a soft trunk robot operating within its workspace. The main idea of this method is to divide the workspace of the controlled trunk robot into several sub-workspaces where the most suitable gains are applied correspondingly in every single sub-workspace. As a result, it becomes feasible to control the trunk by gain scheduling when crossing from one sub-workspace to another as well as considering its dynamic characteristics. The derivation of the method is presented accordingly. In the end, the proposed method is validated by experimental testing with promising results provided afterwards.