Imitation-Learning-in-bipedal-robots

• publication

• Requirements

  • Programming languages:
    • Python
      • packages
        • PyBullet
        • Numpy
        • PyTorch
        • OpenCV
        • Matplotlib
        • Pandas
        • Gym

• Models used

  • Lightweight OpenPose

• How to Run?

  • clone the Repository
  • download the pretrained model for lightweight openpose here.
  • Install the OpenAi gym environment from repository Bipedal-v0
  • run Part_1/imitation_p_1.py to create data folder and save the data in folder.
  • For feed forward network run Part_2/FNN_network.py
  • For Convolutional neural network run Part_2/CNN_network.py
  • both files will plot loss values plot for training

• Series of Experiments

  • Simple movements are tested only where limbs movement are not fast.
  • Cases
    • Adult Male normal walking
    • Adult Female normal walking
    • Adult Male long stride walking
    • Adult Female long stride walking

• imitation_p_1.py will generate joint data angles from the video and save them in a .txt file for each joint, running FNN_network.py and CNN_Network.py will train the model and plot its loss, by running the function data_plot() in Part_1/UTILS.py will generate a plot which compare the original data and smoothed data from savitzky_golay filter and save the figure as smoothed_data.png.