Monte Carlo Tree Search Tutorials

Tutorial notebooks for the IROS 2026 Workshop on Search Algorithms for Robot Learning.

Table of contents

Tutorial	Code
Grid World	Jupyter Notebook Google Colab
Continuous Navigation	Jupyter Notebook Google Colab
Spectral Quadrotor (2D)	Jupyter Notebook Google Colab
Spectral Quadrotor (3D)	Jupyter Notebook Google Colab

Bellmax and LQRax

Bellmax is a compact Monte Carlo tree search solver built with JAX. It provides customizable actions, dynamics, rewards, and search policies, with built-in tree management and value backups.

LQRax is a differentiable continuous-time LQR solver built with JAX, adapted from LQRax. It provides feedback control through backward integration of the Riccati equations. It can be used with Bellmax to combine tree search with feedback control for planning with nonlinear systems.

License

Distributed under the GNU General Public License v3.