MultiRefLfD-TPGP

Learning Multi-Reference Frame Skills from Demonstration with Task-Parameterized Gaussian Processes (TPGP)

DOI | PDF

Installation

Clone the repo

git clone https://github.com/AshrithSagar/MultiRefLfD-TPGP
cd MultiRefLfD-TPGP

Install uv

Install uv, if not already. Check here for installation instructions.

It is recommended to use uv, as it will automatically install the dependencies in a virtual environment. If you don't want to use uv, skip to the next step.

TL;DR: Just run

curl -LsSf https://astral.sh/uv/install.sh | sh

The dependencies are listed in the pyproject.toml file.

Install the package in editable mode (recommended):

# Using uv
uv pip install -e .

# Or with pip
pip install -e .

Additional config

uv tool install bump-my-version

Usage

Warning

WIP

Use the lfd module as a library.

import lfd

D0, _ = lfd.utils.load_data_with_phi("s")

fdset = lfd.utils.transform_data(D0)
P = lfd.alignment.computeP(fdset)
D0_star = lfd.alignment.align_demonstrations(fdset, P)

lfd.alignment.plot_keypoints(fdset, P)
lfd.alignment.plot_alignments(fdset, D0_star, P)

X = lfd.utils.transform_data(D0_star)

Run scripts directly using uv run.

uv run lfd/run.py

References

• Mariano Ramírez Montero, Giovanni Franzese, Jens Kober, and Cosimo Della Santina. Learning Multi-Reference Frame Skills from Demonstration with Task-Parameterized Gaussian Processes. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2832–2839, 2024.

• https://github.com/dfki-ric/movement_primitives

• https://docs.gpytorch.ai/en/latest/examples/04_Variational_and_Approximate_GPs/SVGP_Regression_CUDA.html

• https://docs.gpytorch.ai/en/latest/examples/04_Variational_and_Approximate_GPs/SVGP_Multitask_GP_Regression.html

• https://docs.gpytorch.ai/en/latest/examples/04_Variational_and_Approximate_GPs/Natural_Gradient_Descent.html