This is the repository for the paper "Learning Flexible and Reusable Locomotion Primitives for a Microrobot". More information can be found on our website here. Included are demos for running the experiments laid out in the paper.

Tested and maintained for Python 2.7.12/3.5.2.

Before installing the repo, there are two dependencies that need to be set up manually.

• V-REP, an open-source robotics simulator used to run the experiments (the limited version works if you can't access the educational pro version).
• Opto, a package that implements several of the optimization algorithms used.

To install the remaining dependencies, we recommend cloning into the repo and installing the libraries using pip:

git clone https://github.com/bhyang/microrobot-locomotion.git
cd microrobot-locomotion
pip install -r requirements.txt

Before running any of the experiments, make sure V-REP is open (see the V-REP documentation for troubleshooting issues with installation/booting). Scenes are automatically loaded and can be found in scenes/. The default simulator settings should work fine, but check that the following settings are correct:

• Physics engine: Bullet 2.78
• Time step: 50 ms

To test the single-objective optimization for walking speed only, run:

python normal.py

To run the multi-objective optimization taking into account walking speed and energy efficiency, run:

python moo.py

To run the multi-objective optimization for unbounded gait discovery, run:

python discovery.py

To run the inclination optimization, run:

python incline.py

To run the turning optimization, run:

python turning.py

If you find this code useful, please support us by citing our paper:

Yang, B.; Wang, G.; Calandra, R.; Contreras, D.; Levine, S. & Pister, K. Learning Flexible and Reusable Locomotion Primitives for a Microrobot IEEE Robotics and Automation Letters (RA-L), 2018

@Article{Yang2018,
  Title                    = {Learning Flexible and Reusable Locomotion Primitives for a Microrobot},
  Author                   = {Brian Yang and Grant Wang and Roberto Calandra and Daniel Contreras and Sergey Levine and Kristofer Pister},
  Journal                  = {IEEE Robotics and Automation Letters (RA-L)},
  Year                     = {2018},
  Doi                      = {10.1109/LRA.2018.2806083},
}