Ant Environment

The Ant environment is a high-dimensional continuous control task where a four-legged robot (ant) must learn to walk and navigate through its environment. The goal is to maximize forward velocity while maintaining stability. This environment is commonly used for benchmarking reinforcement learning algorithms. See the Gymnasium documentation for more details.

Ant Animation

The Ant Environment. GIF courtesy of the Gymnasium Documentation.

Tunable Parameters

Parameter

Description

Default Value

gravity

Global gravitational acceleration

[ 0. 0. -9.81]

torso_mass

Mass of the torso

0.32724923474893675

floor_friction

Friction coefficient of the floor

1.0