LiPo: A Lightweight Post-optimization Framework for Smoothing Action Chunks Generated by Learned Policies
pip install action-lipo
-
Clone the repository:
git clone https://github.com/lab-dream/lipo.git
-
Navigate to the cloned directory:
cd lipo -
Install the package:
pip install -e .
-
Visualize lipo: lipo_visualization.ipynb
-
Benchmark:
python lipo_benchmark.py --save media/lipo_benchmark.png
| Implementation | Avg [ms] | Std [ms] | Mean |Δ| vs LiPo | Max |Δ| vs LiPo |
|---|---|---|---|---|
| LiPo | 7.207 | 1.112 | 0.000000 | 0.000000 |
| OSQP QP ⚡ | 0.553 | 0.119 | 0.000740 | 0.005626 |
LiPo can also be used within ROS 2 pipelines. See: https://github.com/lab-dream/lipo_ros2
