This project aims at generating walking locomotion of legged robot autonomously. The training data were generated on the open-source robotic platform Spyndra.
- data: recored on real robot. Suggested structure:
- data - date - gait type - data type
- preprocessing: jupyter notebooks explains each step clearly.
- ml: actual machine learning implementation (keras and sklearn)
- simulation: comparison between real measurement and simulation
- python 2.7
First, download the repository.
$ git clone https://github.com/roboticistYan/Spyndra-ROS-Simulation
Place the collected data into "data" directory.
$ mv YOUR_DATA ~/Spyndra-Gait-Learning/data
Launch the jupyter notebook from downloaed repository.
$ cd ~/Spyndra-Gait-Learning $ jupyter notebook
If your dataset is generated from a single gait, you can verify repeatibility by running the repeatibility analysis notebook.
If your dataset is generated from random gaits, you can run gait feature notebook to gather global paramter dataset.
This work is licensed under MIT License.