Optimising Soft Robot Designs through an Integrated Environment

摘要

Simulation-driven optimisation is increasingly utilised in the design and control development of soft robots and actuators. However, setting up such an optimisation pipeline is complex, often requiring the integration of multiple software tools and algorithms, which can compromise robustness.To address these challenges, we propose a single integrated environment that allows for the flexible description of soft robots and actuators. Our system robustly handles geometry generation, meshing, simulation, and optimisation.We achieve this by using implicit geometry shape functions and voxelisation to create tetrahedral meshes, followed by Extended Position-Based Dynamics (XPBD) to simulate soft materials. As XPBD lacks physical constants, we use an evolutionary optimisation algorithm to calibrate simulation parameters to real-world behaviour and assess how geometry and voxel count affect simulation accuracy. Once calibrated, we find these parameters enable accurate simulations of more complex geometries.Finally, we validate the effectiveness of our integrated environment by optimising a cylindrical soft actuator, demonstrating its potential as an optimisation platform for the field of soft robotics.