RLtools

A C++ deep reinforcement learning library for continuous control, offering algorithms such as TD3, PPO and SAC with CPU and CUDA examples.

Made by rl-tools.

Licenses

Code: MIT (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (MIT) at commit b32d998 (opens in a new tab), on 22 Oct 2026.
Commercial use: Yes

Implements TD3, PPO, multi-agent PPO and SAC, with example environments including Pendulum, racing car, MuJoCo Ant and acrobot.
Includes a Zoo of trained examples built with a single g++ command, run from a seed argument with a browser visualizer.
Can use OpenBLAS or Apple Accelerate to speed up training, as described in the quick start.
Provides Python bindings distributed through PyPI as rltools, usable with Gymnasium environments.
Supports inference and training on iOS and teensy, and inference on Crazyflie, ESP32 and PX4 platforms.
Comes with documentation as C++ notebooks that can be run locally in Docker or through Binder and Colab.

Written by Robotheca from the project's README (opens in a new tab) at commit b32d998.

Difficulty: Advanced
Hardware it needs, as the project states it: A computer with a C++17 compiler; benchmarks mention a 2020 MacBook Pro (M1), an M3 and a Zen 5 Ubuntu machine. CUDA is used for some examples; embedded inference targets microcontrollers such as Crazyflie, ESP32, teensy and PX4 boards.
How to start: The project's website (opens in a new tab)

Checked on GitHub on 22 Oct 2026.

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.