Timezone: »

Context-aware Dynamics Model for Generalization in Model-Based Reinforcement Learning
Kimin Lee · Younggyo Seo · Seunghyun Lee · Honglak Lee · Jinwoo Shin

Wed Jul 15 08:00 AM -- 08:45 AM & Wed Jul 15 07:00 PM -- 07:45 PM (PDT) @ None #None

Model-based reinforcement learning (RL) enjoys several benefits, such as data-efficiency and planning, by learning a model of the environment's dynamics. However, learning a global model that can generalize across different dynamics remains a challenge. To tackle this problem, we decompose the task of learning a global dynamics model into two stages: (a) learning a context latent vector that captures the local dynamics, then (b) predicting the next state conditioned on it. In order to encode dynamics-specific information into the context latent vector, we introduce a novel loss function that encourages the context latent vector to be useful for predicting both forward and backward dynamics. The proposed method achieves superior generalization ability across various simulated robotics and control tasks, compared to existing RL schemes.

Author Information

Kimin Lee (UC Berkeley)
Younggyo Seo (KAIST)
Seunghyun Lee (KAIST)
Honglak Lee (Google / U. Michigan)
Jinwoo Shin (KAIST)

More from the Same Authors