Timezone: »

Bridging Imagination and Reality for Model-Based Deep Reinforcement Learning
Guangxiang Zhu · Minghao Zhang · Honglak Lee · Chongjie Zhang

Mon Dec 07 09:00 PM -- 11:00 PM (PST) @ Poster Session 0 #86

Sample efficiency has been one of the major challenges for deep reinforcement learning. Recently, model-based reinforcement learning has been proposed to address this challenge by performing planning on imaginary trajectories with a learned world model. However, world model learning may suffer from overfitting to training trajectories, and thus model-based value estimation and policy search will be prone to be sucked in an inferior local policy. In this paper, we propose a novel model-based reinforcement learning algorithm, called BrIdging Reality and Dream (BIRD). It maximizes the mutual information between imaginary and real trajectories so that the policy improvement learned from imaginary trajectories can be easily generalized to real trajectories. We demonstrate that our approach improves sample efficiency of model-based planning, and achieves state-of-the-art performance on challenging visual control benchmarks.

Author Information

Guangxiang Zhu (Tsinghua university)
Minghao Zhang (Tsinghua University)
Honglak Lee (Google / U. Michigan)
Chongjie Zhang (Tsinghua University)

More from the Same Authors