SORS: A Modular, High-Fidelity Simulator for Soft Robots

IEEE RoboSoft 2026

Manuel Mekkattu*1, Mike Y. Michelis*1,2, Robert K. Katzschmann1

* Equal contribution 1 Soft Robotics Lab, D-MAVT, ETH Zurich, Switzerland 2 ETH AI Center, ETH Zurich, Switzerland

Soft Robotics Lab logo ETH Zurich logo
arXiv PDF Code Citation

SORS (Soft Over Rigid Simulator) is a modular, energy-based finite element framework for soft robotics. It combines nonlinear material modeling, constrained contact handling, and extensible interfaces for energies, forces, and constraints to support high-fidelity sim-to-real workflows.

Abstract

The deployment of complex soft robots requires simulation frameworks that not only capture interactions between different materials, but also translate accurately to real-world performance. Soft robots pose unique modeling challenges due to their large nonlinear deformations, material incompressibility, and contact interactions, which complicate both numerical stability and physical accuracy. We present SORS (Soft Over Rigid Simulator), a versatile, high-fidelity simulator designed to handle these complexities for soft robot applications. Our energy-based finite element framework allows modular extensions, enabling the inclusion of custom-designed material and actuation models. To ensure physically consistent contact handling, we integrate a constrained nonlinear optimization based on sequential quadratic programming, allowing for stable and accurate modeling of contact phenomena. We validate our simulator through a diverse set of real-world experiments, which include cantilever deflection, pressure-actuation of a soft robotic arm, and contact interactions from the PokeFlex dataset. In addition, we showcase the potential of our framework for control optimization of a soft robotic leg. These tests confirm that our simulator can capture both fundamental material behavior and complex actuation dynamics with physical fidelity. Our approach provides a tool for prototyping next-generation soft robots, filling the gap of extensibility, fidelity, and usability in the soft robotic ecosystem.

Overview

Soft robot simulation needs to balance physical fidelity, extensibility, and practical usability. SORS addresses this with a modular architecture and an energy-minimization backbone that supports materials, actuation, and contact within one consistent pipeline.

  • Modular FEM architecture with minimal physical interfaces: energies, forces, constraints.
  • Constraint-aware optimization via SQP for stable handling of nonlinear contact scenarios.
  • Validated sim-to-real pipeline across cantilever, PokeFlex, soft-arm, and muscle-actuated leg tasks.
SORS overview and application domains
SORS supports extensible modeling across key physical domains and scales from benchmark systems to high-DoF soft robots.

Methods at a Glance

Energy-Based FEM + SQP

Dynamics are solved through constrained energy minimization. At each iteration, SQP solves a local QP to robustly handle nonlinearities, equality constraints, and contact inequalities.

Three Core Interfaces

The framework decomposes physics into energies, forces, and constraints, enabling clear extension points without invasive changes to simulator internals.

Extensible Research Platform

New material laws, pressure/muscle actuation, and contact formulations can be integrated consistently, with a C++ back-end and Python front-end for optimization workflows.

SORS framework architecture
Element-level energy evaluation and global system assembly with extensible interfaces for custom physics.

Results

Cantilever

Marker-based system identification across 17 trajectories achieved 4.98 ± 1.32 mm error with accurate oscillation amplitude and phase matching.

PokeFlex

Contact-rich deformation tracking on volumetric reconstructions reached 6.94 ± 1.66 mm mean Chamfer distance over 10 trajectories.

Soft Arm

Pressure-actuated arm calibration over 6 chamber activations reached 2.53 ± 1.16 mm marker error, even without explicit fiber reinforcement modeling.

Muscle-Actuated Leg

A custom optimization workflow identified actuation timing/strength for jumping, reaching a maximum height of 0.463 m.

Experiment nDoF Trajectories Error Metric (mm)
Cantilever 564 17 4.98 ± 1.32
PokeFlex 3813-5184 10 6.94 ± 1.66
Soft Arm 3249 6 2.53 ± 1.16

Paper Examples

Cantilever simulation animation

Cantilever (Simulation) dynamic simulation rollout.

Cantilever sim-to-real plot

Cantilever (Real Data) sim-to-real displacement benchmark.

PokeFlex simulation sequence

PokeFlex (Simulation) matched simulation sequence.

PokeFlex real-world sequence

PokeFlex (Real Data) real-world deformable interaction sequence.

Soft arm simulation animation

Soft Arm (Simulation) simulation behavior under actuation.

Soft arm sim-to-real comparison

Soft Arm (Real Data) pressure-actuated sim-to-real matching.

Hopping leg simulation animation

Muscle Leg (Simulation) optimized locomotion rollout.

Hopping leg optimization plot

Muscle Leg (Real Data) optimized actuation and jump trajectory.

Known Limitations

  • Contact is currently frictionless (no rolling or tangential friction).
  • GPU acceleration is not yet supported (CPU/OpenMP execution).
  • Contact handling primarily supports rigid primitives; general mesh-mesh contact is limited.
  • No self-collision handling for soft bodies.
  • Meshes are assumed to be clean and well-conditioned (no automatic repair/remeshing).

Citation

@article{mekkattu2025sors,
  author  = {Mekkattu, Manuel and Michelis, Mike Y. and Katzschmann, Robert K.},
  title   = {SORS: A Modular, High-Fidelity Simulator for Soft Robots},
  journal = {arXiv preprint arXiv:2512.15994},
  year    = {2025},
  url     = {https://arxiv.org/pdf/2512.15994}
}