This repo contain the example data generation, training and deployment code for the ICRA 2024 paper "DiPPeR: Diffusion-based 2D Path Planner applied on Legged Robots"
The data generation is structured as a ROS1 package, containing a randomised Maze map generator and a randomised 2D trajectory generator here
alternatively a pre-generated dataset can be found on Huggingface here
The training script is in the form of a jupyter notebook
A pretrained version of the Diffusion based 2D planner model can be found on Huggingface here
A example deployment module us currently provided as ROS1 Navigation global planner plugin which can be found here, once built and sourced, the planner can be integrated by setting the "base_global_planner" config for the navigation stack:
base_global_planner: global_planner/Diffusion2DPlanner