PRobabilistically-Informed Motion Primitives (PRIMP)

Python implementation of PRobabilistically-Informed Motion Primitives, a learning-from-demonstration method on Lie group. This work is published in IEEE Transactions on Robotics (T-RO).

Publication: T-RO
Project page: https://chirikjianlab.github.io/primp-page/
MATLAB version (includes more demos): https://github.com/ChirikjianLab/primp-matlab.

Authors

Sipu Ruan, Weixiao Liu, Xiaoli Wang, Xin Meng and Gregory S. Chirikjian

Dependencies

See requirements.txt

numpy
scipy
finitediff: Compute finite difference
movement_primitives: Python library for movement primitives
dtw-python: Dynamic Time Warping for evaluating distance between learned and demonstrated trajectories
roboticstoolbox-python: Robotics toolbox for different operations and visualizations

Features

PRIMP

Class for the proposed PRIMP method, working on Lie groups. The full 6D pose is considered.

Probabilistic Movement Primitives (ProMP)

Wrapper class that calls the library movement_primitives. The learning spaces include:

Only 3D position of the end effector

Installation

We recommend using pip to install the package:

pip install .

Usage

Data preparation for LfD methods

All test files are located in /test folder. To run scripts for LfD methods:

Download the data from Google Drive. All the demonstrated datasets are locataed in /demonstrations folder.
Generate /data folder that stores all demonstration data
Copy all the demonstration sets into the /data folder
Run scripts in /test folder

Using PRIMP for end effector 6D poses

python main_lfd_primp.py

Using ProMP for end effector 3D positions

Main script for ProMP encoding and conditioning

python main_lfd_promp.py

Benchmark script for evaluations

python benchmark_lfd_promp.py

Generated files

After running, 3 files will be generated (stored in /result/${method}_${planning_group}/):

reference_density_${object}_${demo_type}.json: Full information of the learned workspace trajectory distribution
reference_density_${object}_${demo_type}_mean.csv: Stores only the mean, for seeding the STOMP planner
samples_${object}_${demo_type}.json: Random samples from the learned trajectory distribution

Citation

S. Ruan, W. Liu, X. Wang, X. Meng and G. S. Chirikjian, "PRIMP: PRobabilistically-Informed Motion Primitives for Efficient Affordance Learning from Demonstration," in IEEE Transactions on Robotics, doi: 10.1109/TRO.2024.3390052.

BibTex

@ARTICLE{10502164,
  author={Ruan, Sipu and Liu, Weixiao and Wang, Xiaoli and Meng, Xin and Chirikjian, Gregory S.},
  journal={IEEE Transactions on Robotics}, 
  title={PRIMP: PRobabilistically-Informed Motion Primitives for Efficient Affordance Learning from Demonstration}, 
  year={2024},
  volume={},
  number={},
  pages={1-20},
  keywords={Trajectory;Robots;Probabilistic logic;Planning;Affordances;Task analysis;Manifolds;Learning from Demonstration;Probability and Statistical Methods;Motion and Path Planning;Service Robots},
  doi={10.1109/TRO.2024.3390052}}

ChirikjianLab/primp-python