TRACED: Transition-aware Regret Approximation with Co-learnability for Environment Design

Geonwoo Cho, Jaegyun Im, Jihwan Lee, Hojun Yi, Sejin Kim, Sundong Kim,

Gwangju Institute of Science and Technology

Code

Overview

AMPED is a skill-based reinforcement learning (SBRL) framework that effectively balances exploration and skill diversity, enabling generalizable skill acquisition.

Idea

AMPED Overview Diagram

We propose AMPED, a principled framework that unifies exploration and skill diversity via gradient-level balancing and adaptive skill reuse.

Results

1. Minigrid

2. BipedalWalker

2. BipedalWalker

3. Quantitative Results

4. How important is each component of TRACED?

5. Does it generalized well in the extreme environments?