Numerical design optimisation and real-time simulation of soft fluidic robots for medical gripping applications

The use of soft robotics components, particularly in medical engineering, has grown rapidly over the past two decades. A major research focus lies on soft fluidic actuators, whose continuous deformability and favourable force-to-performance ratio enable a wide range of applications. Their mechanical behaviour is highly complex due to nonlinear material properties and large deformations. The Finite Element Method (FEM) enables accurate numerical simulation of this deformation behaviour, thereby reducing costly prototyping while allowing efficient optimisation of geometries and material parameters.

The deformation characteristics can be further enhanced through targeted fibre reinforcement. Since the material behaviour of such fibres is often time- and temperature-dependent, Artificial Neural Networks (ANNs) are well suited as data-driven material models that can be trained directly on experimental data. In addition, complex FE models can be significantly accelerated through surrogate modelling, for example by means of multiscale modelling or local ANN surrogate models.

For practical applications, the actuation pressure must be adapted to a desired deformation trajectory. Owing to the large number of structural degrees of freedom, this relationship is highly complex. A global ANN can efficiently represent these structural dependencies and be deployed as a real-time control algorithm for deformation-dependent actuation and control.

Contact Person
M.Sc. Maximilian Pawlik