From simulation to navigation.
NEXAH lets agents navigate complex systems by discovering stability instead of optimizing rewards.
No reward. No goal.
Agents converge to stability anyway.
Use cases:
- stabilizing power grids
- navigating chaotic systems
- autonomous scientific discovery
Most simulators only tell you what will happen.
NEXAH shows agents how to steer the system toward stability.
No reward function.
No predefined objective.
Agents do not optimize —
they discover stability.
This is not reinforcement learning. This is navigation.
It works with any simulator through a simple adapter layer.
git clone https://github.com/Scarabaeus1033/NEXAH.git
cd NEXAH
pip install -e .python -m nexah demo kuramoto| Feature | What it does | Example use cases |
|---|---|---|
| SVWIS Operators | Local structural rules for detecting and navigating stability | Precise local navigation |
| Adapter Layer | Connect any simulator in minutes | PowerGrid, Kuramoto, Supply Chain, Cell Biology |
| Multi-Agent Navigation | Multiple agents explore the same landscape together | Coordinated stabilization |
| Visual Regime Landscapes | See stability basins and agent paths live | Real-time understanding |
- Builder Lab Demos – ready-to-run examples you can try immediately
- Live Multi-Agent Demo – multiple agents navigating live in a regime landscape
- Adapter Examples – connect your own simulator in minutes
- Discovery Engine – architecture exploration, resilience analysis & structural law discovery tools
- Core Test Suite – 88+ automated tests validating the entire mathematical kernel
→ Extended Documentation & Full Research Details
To better understand how NEXAH operates across different dynamics and systems, here is an overview visual that shows the key components of the framework and how it interacts with different simulation systems.
This visual illustrates how NEXAH transforms dynamical systems into navigable stability landscapes, highlighting critical points, transition zones, and agent navigation pathways.
Current release: v1.0
- kernel navigation engine implemented
- structural graph models operational
- fixpoint solver validated
- stability analysis modules functional
- modular architecture established
If you use NEXAH in research or academic work, please cite:
Hofmann, T.K.R. (2026).
NEXAH: Structural Navigation in Complex Dynamical Systems
GitHub: https://github.com/Scarabaeus1031/NEXAH
Code: Apache License 2.0
Documentation: CC BY 4.0
© 2026 Thomas K. R. Hofmann