Learning Latent Dynamics for Planning from Pixels
Deep Planning Network (PlaNet), a model-based agent that learns the environment dynamics from pixels and chooses actions through online planning in a compact latent space. To learn the dynamics, we use a transition model with both stochastic and deterministic components and train it using a generalized variational objective that encourages multi-step predictions. PlaNet solves continuous control tasks from pixels that are more difficult than those previously solved by planning with learned models.
The open source code is available here --> https://github.com/google-research/planet <--
This repo contains the source for the article.
Article
draft.md - main text of the article, in markdown.
draft_appendix.md - appendix, in markdown.
draft_bib.html - the citations.
draft_header.html - start of the document
index.html - generated, don't edit this file.
Instructions to Build and Test
git clone https://github.com/planetrl/planetrl.github.io.git
cd planetrl.github.io
npm install
chmod +x ./bin/*Modify text by editing draft.md -- this is where all of the content exists.
Appendix content goes in draft_appendix.md. Add bib entries to draft_bib.html.
Run ./bin/make to build document into index.html (which are identical).
Run python -m http.server to serve on the base directory to view draft.html in a local browser for debugging.
To watch all markdown files for changes and then compile them, you can run the following
brew install fswatch
./bin/watch