Learning Heuristic Search via Imitation

---

Official repository containing OpenAI Gym environments, agents and ML models for the CoRL paper Learning Heuristic Search via Imitation

External Dependencies

1. OpenAI gym
2. Numpy
3. Matplotlib
4. TensorFlow
5. tflearn

Getting Started

One you have installed the required external dependencies (favorably in a virtualenv), you need to execute the following steps in order to get started with the examples.

• Create a meta folder for the project mkdir ~/heuristic_learning
• Get the 2D planning datasets: git clone [email protected]:mohakbhardwaj/motion_planning_datasets.git
• Get the search based planning backend: git clone [email protected]:mohakbhardwaj/planning_python.git
• Get the SaIL repository (this repo): git clone [email protected]:mohakbhardwaj/SaIL.git
• Go to the examples/ folder: cd ~/heuristic_learning/SaIL/SaIL/examples
• Run ./run_generate_oracles_xy.sh which will generate oracles for all the train, validation and test datasets in the motion_planning_datasets folder
• Run ./run_sail_xy_train.sh to train a heuristic for one of the datasets (you can specify the dataset you want inside the script). This runs SaIL for 10 iterations by default. For more information on the rest of the parameters used see the file sail_xy_train.py

Contact

For more information contact Mohak Bhardwaj : [email protected]