Optimization of soft manipulator simulation using a particle swarm approach

Abstract

Simulation is a valuable tool in soft robotics, but it is essential that it accurately reflects reality. Calibration with real-world data ensures that simulation parameters correlate with physical conditions. Data from a real soft manipulator was collected, compared to a simulated environment to optimize it. A particle swarm optimization was employed to find the optimal Young’s modulus that aligns real-world and simulated conditions. This study investigates the behavior of soft manipulators with varying numbers of springs in a simulated environment, using both Euler implicit and Newmark implicit solvers. It also examines how manipulator configurations and solver methods impact Young’s modulus, optimization costs, and solver efficiency. The solver method and number of springs critically impact Young’s modulus and simulation complexity due to their computational characteristics and the non-linearities absorbed by the simplified model. By analyzing setups with one to five springs across these solvers, the study reveals patterns and distinctions useful for future optimization in soft robotics design and simulation.