Final project CS282A: Diffusion Planning

Authors: Carlo Bosio¹, Karim Elmaaroufi², Margarita Geleta², Matteo Guarrera²

¹ UC Berkeley, Department of Mechanical Engineering

² UC Berkeley, Department of Electrical Engineering and Computer Science

Description

This repository contains the code for our project on diffusion models for drone motion planning.

Usage

pip install -r requirements.txt

Execution steps

1. Download the pretrained models from the v0.0.1 release

2. Launch training Run the following, if you have the computational power.

  >> python diffusion.py

If not, we suggest using the pretrained model and just do inference.

If you are using a M1 Mac, please execute any jupyter notebook with the following command to enable MPS:

PYTORCH_ENABLE_MPS_FALLBACK=1 jupyter notebook

3. Inference Run the DiffusionInference notebook and visualize one trajectory.

Simulator

For the simulator installation please follow this link.