Sampling-Based Optimization

Repo for testing sampling-based optimization methods.

Installation

Use conda and install via:

conda env create -f environment.yml
conda activate env_sbo

Directories

• scripts/: Contains the main scripts for running experiments and testing optimization methods.
• srb/: Contains the scripts for Single Rigid Body (SRB) non-linear trajectory optimization.
• models/: Contains various dynamical system models described by an XML file. You can view a model using visualize_model_mujoco.py.
• utils/: Contains utility functions and classes used throughout the codebase. For example, parallel simulation, optimization methods, and splines.
• results/: Directory to store results from experiments. To view results, run view_results.py which will load the data, visualize it, and plot. Just change the experiment and xml path at the top of the file to view different results. Similarly, you can use view_results_srb.py to view results from the SRB optimization.
• logs/: Directory to store tensorboard logs from experiments. View with tensorboard --logdir=./logs.

Usage

All scripts are run from the root directory. The main scripts are located in the scripts/ directory.

If you can't run a particular script, try running it as a module. For example, for scripts/cartpole/cartpole_cem.py, use:

python -m scripts.cartpole.cartpole_cem

instead of:

python scripts/cartpole/cartpole_cem.py

Viewing Results

Optimization Results

Optimization results are saved in results/. To view, run view_results.py which will load the data, visualize it, and plot. Just change the experiment and xml_path at the top of the file to view different results. Similarly, you can use view_results_srb.py to view results from the SRB optimization.

Tensorboard Logs

To view tensorboard logs for optimization progress, run:

tensorboard --logdir=./logs

and open the provided URL in your browser (e.g., http://localhost:6006). To enable logging for a task, pass a Logger_Config to the optimizer.