Kinematic Modeling of a Miniaturized Configurable Parallel Robot with Soft Joints for Real-Time Control

Reaching a high level of miniaturization and precision can be challenging when using classical mechanical joints. In contrast, soft joints can mimic the kinematic behavior of classical joints while enhancing precision by addressing issues like backlash. However, the accuracy of these joints heavily depends on their mathematical models, typically derived using Finite Element Models (FEM). Unfortunately, FEM models are time-consuming, which significantly complicates their implementation for real-time control. Consequently, the robot's positions must be pre-learned, restricting its ability to accurately reach new positions in real-time. In our study, we demonstrate that cubic soft joints can accurately mimic the kinematics of classical spherical joints, allowing for rapid computation of the robot's kinematic model. Comparisons with experimental results from a parallel miniature robot equipped with cubic soft joints and a configurable platform showed no significant difference between the proposed approximate model and more complex models like fine-meshed FEM. However, the simplified model operates thousands of times faster, which facilitates its implementation for real-time control. This breakthrough enables the modeling and design of complex miniaturized robot architectures, fostering new capabilities for dexterous manipulation at small scales.