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

• Publication: T-RO
• Project page: https://chirikjianlab.github.io/primp-page/
• Python implementation is available here.

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

• Robotics Toolbox (RVC 2nd edition: RTB10+MVTB4 (2017))
• PbD Library
• Orientation-KMP

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 (Only put folders inside /demonstrations into /data)
• (Optional) To generate data trials for real-world experiments, please also download datasets from /experiments folder (Put the whole folder).
• Run scripts in /test folder

After data preparation, the structure of /data folder should look like

.
└───data
│   └───panda_arm
|   |   └───real
|   |   |   └───trajectory
|   |   |   |   └───...
|   |   └───simulation
|   |   |   └───...
│   └───lasa_handwriting
|   |   |   └───pose_data
|   |   |   |   └───...
│   └───experiments
|   |   └───...

To run Orientation-KMP method,

• Clone from Orientation-KMP Repository
• Add correct paths of functions

addpath path-prefix/pbdlib-matlab/demos/m_fcts/
addpath path-prefix/robInfLib-matlab/fcts/

DTW is required to evaluate method performance. The source C code is /src/util/dtw_c.c, which needs to be generated as a .mex file.

• Go to /src/util
• Mex the C code for DTW

mex dtw_c.c

• Plot LfD dataset /test/demo_lfd_dataset.m: Plots all demonstration trajectories and store .png files in /test folder.

• Encode demonstrations and condition on via poses /test/demo_primp_lfd.m
• LfD for scooping task /test/demo_primp_lfd_scooping.m
• Encode demonstrations with and without GORA /test/demo_primp_lfd_gora.m
• Adaptation: Pass through via poses with uncertainties /test/demo_primp_condition_via_poses.m
• Adaptation: Equivariance on the change of viewing frame /test/demo_primp_change_view.m
• Adaptation: Fusion with workspace density /test/demo_primp_lfd_fusion_wd.m

• Encode demonstrations and condition on via poses /test/demo_kmp_lfd.m

• Generate random via-point poses data for benchmarks (Required before running benchmark scripts) /test/generate_benchmark_trials.m
• Benchmark LfD for PRIMP /test/benchmark_lfd_primp.m
• Benchmark LfD for PRIMP with storage of learned trajectory distribution /test/benchmark_lfd_primp_trajectory.m
• Benchmark LfD for Orientation-KMP /test/benchmark_lfd_kmp.m
• Benchmark for PRIMP on synthetic data between SE(3) and PCG(3) formulation /test/benchmark_primp_se_pcg.m
• Benchmark for PRIMP on single demonstration /test/benchmark_lfd_primp_single_demo.m
• Benchmark for PRIMP on equivariant change of view /test/benchmark_lfd_primp_change_view.m
• Qualitative comparisons among LfD methods for extrapolation cases /test/comparison_lfd_extrapolation.m

• Ablation study for GORA as pre-process /test/ablation_primp_gora.m
• Ablation study for fusion with robot-specific workspace density /test/ablation_primp_wd.m

• Generate reference trajectory using PRIMP for real-world experiments /test/generate_real_task_trials.m

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}}