@realyunlong on Backlist

Robotics RL is often imitation learning in disguise

Reward shaping, curricula, initialization, and environment design can smuggle human demonstrations into reinforcement-learning systems indirectly