Content

• Description/Abstract
• How-to-RUN
• How does it work?
• Additional

How-to-RUN

Do you want to try our code? Below, you can find a step-by-step guide that will show you how to do it.

Note that I use Linux, some of the steps may differ depending on OS.

Requirements

• Python interpreter python3
  • Install torch from link: PyTorch
  • Install other required packages using pip install -r requirements.txt

Execution

• Run grid_maker.py using python grid_maker.py
• Run data_driver.py using python data_driver.py --data grid_maker --num_samples 10000

How does it work?

Do you want to know how our code works? Below, you can find an overview of how it works.

Grid Dataset Generation

• Using tools in grid_maker.py we can generate a NxN grid, $\text{N}\in \mathbb{Z}$. Grid is empty, filled with 0's. Then we randomly generate initial positions for 2 objects. Given the grid, and object positions, we now can generate their actions. It is possible to generate a colliding actions, and non-colliding.
• data_driver.py just wraps it up nicely and creates a .csv dataset.

Data Pre-Processing

• Not Implemented yet