Quadruped Walking With Reinforcement Learning

This repository contains code to train an RL agent for different tasks. The abstract QuadrupedEnv class can be extended to create custom environments for different tasks. Every implemented task includes a environment class, a training script, and a rendering script.

Implemented Tasks

Walking

• Environment
• Training Script
• Rendering Script

Recurrent Walking

Walking trained with a recurrent policy (PPO-LSTM).

• Environment
• Training Script
• Rendering Script

Jumping

• Environment
• Training Script
• Rendering Script

ZMP Walking

Walking with awareness of the Zero Moment Point (ZMP).

• Environment
• Training Script
• Rendering Script