/irbfn

Code for IROS 2023 paper: Differentiable Trajectory Generation for Car-like Robots by Interpolating Radial Basis Function Networks

Primary LanguageJupyter NotebookMIT LicenseMIT

Interpolating Radial Basis Function Networks

Code for IROS 2023 paper: Differentiable Trajectory Generation for Car-like Robots by Interpolating Radial Basis Function Networks Trajectory

Prerequisite

  1. NVIDIA driver version 520.61.5+ is required.

Installation

  1. Clone this repo, then cd irbfn
  2. Update submodule. git submodule sync && git submodule update --init --force
  3. Build docker image from Dockerfile:
sudo docker build -t irbfn -f Dockerfile .
  1. Run the docker container:
sudo ./run_container.sh

Run evaluation

python3 evaluate.py

Re-run training

  1. Download MPC training dataset here
  2. Download training dataset. See instructions here.
  3. Run train script python3 train.py. Updated config and checkpoint files should be saved to ckpts/ and configs/.