SoRoSim — Implicit Routing

Differentiable simulation of soft continuum manipulators with implicit internal cable routing, built on the SoRoSim (Soft Robotics Simulator) toolbox. Given a set of cable entry/exit ports, the cable's path through the manipulator is solved for implicitly (as part of the static equilibrium) rather than prescribed, and the whole pipeline is differentiable end-to-end (analytical Jacobians throughout).

Repository layout

ImplicitRouting/
├── SoRoSim_Toolbox/                        Shared library - one copy used by every project below
│   ├── Basic_functions/                    Screw-theory / Lie-group primitives, MEX-accelerated kinematics
│   ├── SorosimLink_files/                  SorosimLink class (rigid/soft link definition)
│   ├── SorosimLinkage_files/                SorosimLinkage class (statics, dynamics, plotting, actuation)
│   └── SorosimRod_files/                   SorosimRod class (Cosserat rod strain bases)
│
├── Single Section Single Cable ParID/       Parameter identification: fits section stiffness (c1..c5)
│                                            from marker-tracked routing experiments
├── Single Section Single Cable Simulation/  Implicit-routing statics demo, mesh-convergence and
│                                            solver-tolerance studies (no experiment data required)
├── Single Section Three Cables IKS/         Inverse kinetostatics: solves cable inputs to achieve a
│                                            sequence of desired tool-tip pointing directions
└── Two Sections Two Cables Simulation/      Two-section, two-cable manipulator statics demo

Every project folder is self-contained (its own driver script(s), model .mat files, Custom/ actuation hooks, startup.m, and ReadMe.txt) and only depends on the single shared SoRoSim_Toolbox/ next to it — see each project's own ReadMe.txt for its specific workflow.

Requirements

• MATLAB (developed/tested on Windows; not verified on macOS/Linux)
• Optimization Toolbox (fmincon, fsolve)
• MATLAB Coder — only needed to (re)build the MEX-accelerated kinematics functions, see below

Getting started

1. Clone the repository.
2. cd into whichever project folder you want to run (e.g. Single Section Single Cable Simulation/) and run startup.m. This adds the shared SoRoSim_Toolbox/ (resolved relative to the project folder, so it works regardless of MATLAB's current directory) plus that project's own Custom/ folder to the path.
3. Run one of that project's driver scripts (e.g. driver.m). See the project's own ReadMe.txt for what's available and its user-option flags.

Building the MEX files (converttoMEX.m)

The kinematics functions used in the inner solver loop are pre-compiled to MEX for speed. The compiled *.mexw64 binaries are not committed to this repository (they're platform- and MATLAB-version-specific build artifacts — see .gitignore), so the first time you use this repository, or after a MATLAB update / on a new machine, you need to build them yourself:

cd SoRoSim_Toolbox/Basic_functions
converttoMEX

This uses MATLAB Coder (codegen) to compile the functions listed inside converttoMEX.m and drops the resulting *_mex.mexw64 files alongside the .m sources in Basic_functions/, where every project's startup.m will find them automatically. If any driver script errors with something like Undefined function 'RigidJointKinematics_mex' or a MEX-related error, this is almost always the fix — re-run converttoMEX.m.

Citation

If you use this code in your research, please cite:

@article{mathew2026irouting,
  author   = {Mathew, Anup Teejo and Alkayas, Abdulaziz Y. and Alshehhi, Ahmed A. S. and Adamu, Yusuf A. and Taha, Tarek and Renda, Federico},
  title    = {A Differentiable Framework for Hollow Tendon-Driven Continuum Robots With Implicit Internal Routing},
  journal  = {IEEE Transactions on Robotics},
  year     = {2026},
  doi      = {10.1109/TRO.2026.3714660},
  keywords = {Continuum robot, differentiable simulation, optimization, tendon-driven continuum robots}
}

Video

Supplementary Video

Contact

Anup Teejo Mathew 📧 anup.mathew@ku.ac.ae · anupteejo@gmail.com