Skip to content

Latest commit

 

History

History
86 lines (74 loc) · 3.2 KB

README.md

File metadata and controls

86 lines (74 loc) · 3.2 KB
Raw

Central Pattern Generator For Quadruped Robot

Creator - Damian Bemben

The following files implement a central pattern generator for a quadruped robot in PyBullet as well as a method for parsing and measuring quadruped data based on the Dynamic Similarity Hypothesis.

Acknowledgments

Original Van Der Pol Oscillator Code - Andrew Walker

PyBullet - Erwin Coumans

Requirements

In order to run this code, the following libraries are required. The libraries can be installed by running the following command:

pip install -r pipmodules.txt

pybullet numpy matplotlib scipy sklearn pandas

Example Run

In order to run an example of the Laikago in a graphic environment with a stable gait, run the following command.

laikago_GUI.py 40 0.006 0 x 10 10

laikago.py

Takes in 6 parameters for running, outputs a file containing mean and standard deviation values for experiments:

  1. Force - (20 - 100)
  2. Oscillator Timesteps (0.01 - 0.002)
  3. Gait - (0 - Walking, 1- Trotting, 2- Bounding)
  4. Folder to save to - (Any String ("all" = default))
  5. Foot Angle - (5-20)
  6. Hip Angle - (5-20) Runs for 10000 iterations with 10 repetitions, prints out results in python file, with the following format:
Oscillator Step: ()
Max Force: ()
Gait: ()
Leg Rotation: ()
Hip Rotation: ()
Velocity Mean: Velocity SD
Froude Mean: Froude SD
Distance Mean: Distance SD
Cost Of Locomotion Mean: Cost of Locomotion SD
Time Period Mean: Time Period SD

An example of parsing can be found in plot_main_results.py -> parse_all()

Variations on laikago.py

Although laikago.py runs the majority of experiments, laikago_structural.py can be used to find successful gaits. This uses the same parameters, but only runs for 10000 iterations, and only outputs the following format:

(Distance)
(Froude Number)

ttest.py runs for 10000 iterations and produces a paired t-test of values.

Experiments

Experiments can be seen by running plot_main_results.py. This parses a folder and returns all experiments. Current set of experiments are the following:

print_percentages() - Prints percentage of values below froude number values of 0.4

oscillator_vs_froude_bar() - Prints a bar chart of average oscillator time-step vs froude number ' plot_3d() - plots a 3d graph of all combinations of values plot_distribution() - plots distributions found in dissertation

plot_froude() - plots froude number graph found in dissertation

Misc

van.py

Produces a single uncoupled van der pol oscillator

van_coupled_rt.py

Produces a real-time graph of coupled van der pol oscillators

dynamic_similarity.py

Produces an example of dynamic similarity