Update README.md

README.md

@@ -2,6 +2,8 @@
 This repository provides a demonstration of imitation learning using a diffusion policy. The implementation is adapted from the official Diffusion Policy [repository](https://github.com/real-stanford/diffusion_policy).
 
 ## Result
+The animation shows the denoising process of the diffusion policy predicting future trajectory and applying actions.
+
 <img src="assets/result_anim.gif" alt="drawing" width="380"/> <img src="assets/result_plot.png" alt="drawing" width="370"/>
 
 
@@ -14,4 +16,9 @@ The program was developed and tested in the following environment.
 - `diffusers==0.18.2`
 - `torchvision==0.14.1`
 - `gdown` (to download pre-trained weights)
-- `joblib` (format of training data)
+- `joblib` (format of training data)
+
+## Diffusion policy
+The policy takes 1) the latest N step of observation (position and velocity) and 2) encoding of obstacle information (7x7 grid with obstacle radius as values) as input and outputs N steps of actions (future position and future velocity).
+
+<img src="assets/model_input.jpg" alt="drawing" width="480"/>