/Embed2Control

Primary LanguageJupyter Notebook

Embed to Control

This is re-implementation of "Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images", NIPS, 2015. known as E2C. We tested on the pendulum environment. For details please refer to our report.

E2C code is from E2C pytorch implementation and the changes are as follows:

  • We changed the consistency loss term. There is a typo in the original paper.
  • We changed the code of generating the images.

iLQR code is from ilqr_pendulum and the changes are as follows:

  • We integrated the E2C model.
  • We added anlaysis plots.

This is joint work with Kihong-cyber.