Coupling field simulation of soft capacitive sensors towards soft robot perception

Simulation is a standard tool for robot design, control, and performance analysis. Numerous mature, thoroughly validated methods exist to build fast and reliable simulation models for traditional rigid robotics platforms. However, when it comes to novel soft robotics systems, most existing models concern themselves with the dynamics of the soft bodies during actuation, largely disregarding the sensory system. Simultaneous simulation of sensors and actuators is essential to establishing a realistic and accurate robot model in the virtual environment. This article proposes a pipeline to implement coupling field simulation (CFS) of capacitive sensors deployed on a square soft arm and a pneumatic manipulator. The CFS approach can seamlessly integrate mechanical and sensing components, enabling us to understand sensor behavior better by simulating sensor responses to various deformations including bending, inflation, and the combinations of bending, twisting, and elongation. We also demonstrate that CFS is cost-effective manner to acquire a large number of annotated data that can be used for pretraining in soft robot perception tasks through two case studies (i.e., applied force estimation and deformation classification).