How to Train Dynamical Models for Model Predictive Control
Implementation Code and Experimentation Results for MLSS2024 Poster Presentation
Created on February 27|Last edited on March 8
Comment
Videos
Simple Method
videos
Hybrid Method
Learning Curve
Walker
Cheetah
Quadruped
Dog
Humanoid Result
Colab Notebooks
Simple Method
Hybrid Method (TD-MPC)
Note: The following poster describes the problem setting when the reward function is known. However, in the above Colab Notebook, due to limitations of the library used, the reward function is trained using a neural network as well as dynamics.
Contents of Poster
Add a comment