useful_info.txt

# File/folder descriptions

common, include, stl_files   - contain files which are included in the pusher domain

config.py    - Contains general parameters used throughout including for coarse and fine models.

data_generator.py    - generates state data for an environment with an applied action. Applies a first action (not recorded) in order to create realistic object velocities.Then it uses the resulting state as input to which another action is applied. The input state, action, and resulting state are recorded. Also renders the real frames.

generate_plots.sh - It runs the python script for experimental data plotting with many different parameters.

model_weights.h5  - the weights of a trained model that can predict for up to 4 sliders.
		Network configuration: 25(=input),512,256,128,64,24(=output), activation=relu, use_bias=True

parareal_manipulation.py - Implements two coarse models and one fine model for Parareal.Used to generate experimental data.

plot_generation.py - Uses experimental data to generate results: plots, imgae frames, and text files.

run_experiment.sh - Runs the parareal_manipulation.py script with several inputs to generate experimental data.

test_model.py 	  - takes a trained model and test data and computes the error in its predictions.
		Also can render predicted final state vs. correct state.

train_model.py    - takes csv files containing training and testing data to train a model.
Can save the model or its weights to be used or tested, or re-trained with different hyperparameters or a different dataset afterwards. Has the option of using a custom loss function which penalizes object penetration.

# Sample result:

Fig. 2 error plot in Agboh et. al. (CVS 2020) can be found in:

	exp_dataset/1slider_4/parareal_plots/1S_rms_log_xy_error.png

I.e after data generation and plotting / direct data download.

# Further details 
This plot is the xy position error log plot for 1 slider and a control sequence with 4 actions.
Data used to generate the plot is in exp_dataset/1slider_4/parareal_data
Image frames for Parareal with the learned model are found e.g. in:
	exp_dataset/11slider_4/parareal_frames/learned_model/sequence_1/iteration_0
Note that iteratio 0 is the coarse prediction.
Physics simulation time plots are in exp_dataset/11slider_4/time_plots
Max. Object displacements are written to file e.g. exp_dataset/1slider_4/results.txt