Exact Smooth Reformulations for STL Trajectory Optimization

Why this tool?

Signal Temporal Logic (STL) gives spatial-temporal specifications, but enforcing them in trajectory optimization is typically non-smooth and hard to solve.

We propose exact smooth reformulations, enabling off-the-shelf derivative-based solvers (e.g. SNOPT) without sub-optimality issues.

Installation

Tested on Python 3.10 / Ubuntu 22.04. To set up the environment, run:

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

To run MICP baselines, install Gurobi and set up an (academic) license.

To enable pre-commit hooks for code quality checks, run:

pre-commit install

Usage

To run the experiments, check out the folder experiments.

Acknowledgements

This work builds on the stlpy.

BibTeX

If you find this work useful, please consider citing:

@article{han2025exact,
  title={Exact Smooth Reformulations for Trajectory Optimization Under Signal Temporal Logic Specifications},
  author={Han, Shaohang and Verhagen, Joris and Tumova, Jana},
  journal={arXiv preprint arXiv:2511.07375},
  year={2025}
}