Skip to content

Dream to Control: Learning Behaviors by Latent Imagination

Authors: Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi

Published: 2019 (Conference Paper)

Source: International Conference on Learning Representations (ICLR)

Algorithm: Dreamer

arXiv: 1912.01603

Summary

Build on the predecessor work PlaNet and learns an actor-critic model in place of online planning with the cross entropy method.

Abstract

Learned world models summarize an agent's experience to facilitate learning complex behaviors. While learning world models from high-dimensional sensory inputs is becoming feasible through deep learning, there are many potential ways for deriving behaviors from them. We present Dreamer, a reinforcement learning agent that solves long-horizon tasks from images purely by latent imagination. We efficiently learn behaviors by propagating analytic gradients of learned state values back through trajectories imagined in the compact state space of a learned world model. On 20 challenging visual control tasks, Dreamer exceeds existing approaches in data-efficiency, computation time, and final performance.

Tags

  • World models

  • Deep learning

  • Reinforcement learning

  • Latent

  • Imagination

  • Complex behaviors