https://github.com/nepfaff/gcs_planar_pushing

https://github.com/nepfaff/gcs_planar_pushing

Last synced: 10 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/nepfaff/gcs_planar_pushing
• Owner: nepfaff
• Created: 2023-03-24T13:07:19.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-10-07T15:15:38.000Z (over 1 year ago)
• Last Synced: 2025-01-14T06:16:51.714Z (16 days ago)
• Language: Python
• Size: 578 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# gcs_planar_pushing

See the [report](report.pdf) for more information about this project.

## Installation

Clone the repo and execute the following commands from the repository's root.

Install the `gcs_planar_pushing` package in development mode:

```

pip install -e .

```

Install `pre-commit` for automatic black formatting:

```

pre-commit install

```

## Generating GCS demonstrations

The following script can be used to generate demonstrations for training the diffusion policy.

The demonstrations are created using GCS planning and open-loop execution.

```

python scripts/generate_demonstrations.py --config-name demonstrations.yaml \

only_generate_initial_positions=False problem_generator.n_samples=100

```

## Runing a single experiment

Create a config file specifying the experiment in `config` and run it using the following command:

```

python scripts/run_demo.py --config-name basic.yaml

```

where `basic.yaml` should be replaced with your config name.

### Running Diffusion Policy

A pre-trained diffusion policy controller can be run as follows:

```

python scripts/run_demo.py --config-name basic.yaml controller=diffusion_policy \

environment.initial_box_position='[2,0]' environment.initial_finger_position='[3,3]'

```

Running with random disturbances:

```

python scripts/run_demo.py --config-name basic.yaml controller=diffusion_policy \

environment.initial_box_position='[-1,2]' environment.initial_finger_position='[3,3]' \

environment.disturbance_probability_per_timestep=0.005 environment.disturbances_max_number=3

```

## Evaluating controllers on planar cube environment

Example: Evaluating diffusion policy without disturbances:

```

python scripts/evaluate_planar_cube_env.py --config-name evaluate_planar_cube.yaml \

initial_conditions_zarr_path=data/initial_positions_2023-05-13_10-36-08.zarr/

```

where `initial_conditions_zarr_path` is the path to a `zarr` file containing the initial

experiment conditions with the following data format:

- "object_pos" of shape (N,2)

- "robot_pos" of shape (N,2)

Example: Evaluating diffusion policy with disturbances:

```

python scripts/evaluate_planar_cube_env.py --config-name evaluate_planar_cube.yaml \

environment.disturbance_probability_per_timestep=0.005 environment.disturbances_max_number=3 \

initial_conditions_zarr_path=data/initial_positions_2023-05-13_10-36-08.zarr/

```

Example: Evaluating GCS plan without disturbances:

```

python scripts/evaluate_planar_cube_env.py --config-name evaluate_planar_cube.yaml \

initial_conditions_zarr_path=data/initial_positions_2023-05-13_10-36-08.zarr/ \

controller=gcs_open_loop

```