Reset related

PAIRED - Protagonist and Antogonist Adversarial Learning

Optimal Control: Model-Based RL

THE INGREDIENTS OF REAL-WORLD ROBOTIC REINFORCEMENT - reset to scenes that are not in known distribution to increase exploration

Curriculum related

AUTOMATIC CURRICULUM GENERATION FOR REINFORCEMENT LEARNING IN ZERO-SUM GAMES

Asymmetric selfplay

Intrinsic Motivation

Real Robot Learning

Assessing Generalization in Deep Reinforcement Learning

DensePhysNet: Object physics representation

UMP: 6DOF representation

CS285

Synthesizing Dexterous Nonprehensile Pregrasp for Ungraspable Objects