ac31415 / c3m_mod

Learning Certified Control Using Contraction Metric (CoRL 2020)

Learning Certified Control Using Contraction Metric

Pytorch implementation of the CoRL'20 paper "Learning Certified Control Using Contraction Metric", by Dawei Sun, Susmit Jha, and Chuchu Fan.

Requirements

Dependencies include torch, tqdm, numpy, and matplotlib. You can install them using the following command.

pip install -r requirements.txt

Usage

The script main.py can be used for learning the controller. Usage of this script is as follows

usage: main.py [-h] [--task TASK] [--no_cuda] [--bs BS]
               [--num_train NUM_TRAIN] [--num_test NUM_TEST]
               [--lr LEARNING_RATE] [--epochs EPOCHS] [--lr_step LR_STEP]
               [--lambda _LAMBDA] [--w_ub W_UB] [--w_lb W_LB] [--log LOG]

optional arguments:
  -h, --help            show this help message and exit
  --task TASK           Name of the model.
  --no_cuda             Disable cuda.
  --bs BS               Batch size.
  --num_train NUM_TRAIN
                        Number of samples for training.
  --num_test NUM_TEST   Number of samples for testing.
  --lr LEARNING_RATE    Base learning rate.
  --epochs EPOCHS       Number of training epochs.
  --lr_step LR_STEP
  --lambda _LAMBDA      Convergence rate: lambda
  --w_ub W_UB           Upper bound of the eigenvalue of the dual metric.
  --w_lb W_LB           Lower bound of the eigenvalue of the dual metric.
  --log LOG             Path to a directory for storing the log.

For example, run the following command to learn a controller for the 8-dimensional quadrotor model.

mkdir log_QUADROTOR_8D
python main.py --log log_QUADROTOR_8D --task QUADROTOR_8D

Run the following command to evaluate the learned controller and plot the results.

python plot.py --pretrained log_QUADROTOR_8D/controller_best.pth.tar --task QUADROTOR_8D --plot_type 3D --plot_dims 0 1 2
python plot.py --pretrained log_QUADROTOR_8D/controller_best.pth.tar --task QUADROTOR_8D --plot_type error

If you find this project useful, please cite:

@article{sun2020learning,
  title = {Learning certified control using contraction metric},
  author = {Sun, Dawei and Jha, Susmit and Fan, Chuchu},
  booktitle = {Proceedings of the Conference on Robot Learning},
  year = {2020}
}