A Simulation Approach To Soft Robots Based On The Material Point Method

The high flexibility and adaptability make soft robots safely interact with task targets in complex environments.However,the large deformation and complex interaction in soft robot tasks will complicate the analysis and modeling during the structural design procedure of soft robots.To evaluate and verify the structure of the soft robot in time during and after the design procedure,the physical simulation of the soft robot is performed using a method based on the continuum mechanics.But currently used mesh-based methods for soft robot simulation fail in the free boundary coupling problem caused by large deformation and complex interaction as the lack of prior knowledge for correct contact relations.In this thesis,a Material Point Method framework combined with automatic differentiation is proposed and applied to soft robot simulation.The proposed approach can autonomously solve the free boundary coupling problem caused by large deformation and complex interaction.The main innovations include:（1）A computation framework of Material Point Method based on automatic differentiation is proposed,the material models and automatic differentiation are combined to compute the stress,which avoids the difficulty of hand calculation of multidimensional derivatives,and the optimal problem for implicit time integration framework is solved by automatic differentiation as well.（2）Material Point Method is applied to the soft robot simulation,by analyzing and decoupling of the soft robot simulation,the computation framework,material models and driven source module of the simulation program are accomplished,the free boundary coupling problem caused by the large deformation and complex interaction of the soft robot in the simulation is solved,so that the soft robot structural design procedure can take the task loads and environment elements into account.