SORS: A Modular, High-Fidelity Simulator for Soft Robots
A new energy-based framework tackles soft robots' unique challenges—nonlinear deformations, material incompressibility, and contact—with high physical fidelity.
A research team from ETH Zurich, led by Manuel Mekkattu, Mike Y. Michelis, and Robert K. Katzschmann, has introduced SORS (Soft Over Rigid Simulator), a new simulation framework designed to overcome the unique modeling challenges of soft robots. These challenges include large nonlinear deformations, material incompressibility, and complex contact interactions, which have historically complicated numerical stability and physical accuracy in simulation. SORS is built on an energy-based framework utilizing the finite element method, allowing for modular extensions that can incorporate custom material and actuation models. A key innovation is its integration of a constrained nonlinear optimization based on sequential quadratic programming, which ensures physically consistent and stable contact handling—a notoriously difficult problem in soft robotics.
The simulator's capabilities were rigorously validated through a diverse set of real-world experiments. These included fundamental tests like cantilever deflection, pressure-actuation of a soft robotic arm, and contact interactions from the established PokeFlex dataset. Furthermore, the team demonstrated SORS's practical application by using it for control optimization of a soft robotic leg. These tests confirm that the simulator can capture both basic material behavior and complex actuation dynamics with high fidelity. By providing a tool that accurately bridges the simulation-to-reality gap, SORS addresses a critical need for extensible, high-fidelity prototyping in the soft robotics field, potentially accelerating the development and deployment of adaptable robots for healthcare, exploration, and delicate manipulation tasks.
- Built on a modular, energy-based finite element framework that allows for custom material and actuator models.
- Uses constrained nonlinear optimization (sequential quadratic programming) for stable and accurate modeling of contact phenomena.
- Validated with real-world experiments including the PokeFlex dataset and control optimization for a soft robotic leg.
Why It Matters
Provides a validated, high-fidelity tool for prototyping adaptable robots, accelerating R&D in healthcare, exploration, and delicate automation.