Pendulum-v1
A continuous-control benchmark for swinging a pendulum upright and keeping it there.
Published by Farama Foundation
Setup and evaluation
The policy selects a continuous torque, which makes this a useful step after discrete-control examples. The reward accounts for angle, angular speed, and torque, so an agent has to balance accuracy with control effort.
Install with pip install gymnasium and instantiate gymnasium.make('Pendulum-v1'). Compare mean return across the same evaluation seeds and keep the episode limits consistent.
- Framework
- Gymnasium
- Access
- Local Gymnasium environment. No account required.