Few-Shot Preference Learning for Human-in-the-Loop RL

Joey Hejna, Dorsa Sadigh

While reinforcement learning (RL) has become a more popular approach for robotics, designing sufficiently informative reward functions for complex tasks has proven to be extremely…