Update on 13 Nov 2020: This repository has moved! Please check out the latest version here: https://github.com/roahmlab/arm_planning
This repo contains the code for our recent paper: https://arxiv.org/abs/2002.01591
To run this code, you'll need MATLAB R2018b or newer. Also, you will need CORA_2018.
If you want to run the GPU versions of the code, you'll need IPOPT... more detailed instructions for setting up the GPU versions of the code to follow.
All of the ARMTD MATLAB/MEX code has a non-GPU version, but it isn't usually real-time fast. You may have to recompile the CUDA MEX files for your own system, if you're using a GPU.
The armtd_benchmark
folder contains code for running a comparison of our proposed ARMTD method against CHOMP. This requires Ubuntu with MoveIt! and ROS installed. Similarly, everything in the fetch_command_files
folder is what we used to communicate with the Fetch hardware (via ROS in MATLAB, from an Ubuntu machine).
Note, none of the comparison or hardware code is required for just running ARMTD in simulation! You do not need to worry about this stuff unless you really want to replicate everything from our paper and hardware experiments.
Navigating this Repo (start here)
Here's a brief explanation of the different folders here:
- armtd_benchmark: code for comparing armtd to CHOMP through MoveIt!
- fetch_command_files: code for interfacing with the Fetch through MATLAB and ROS
- figures: a couple scripts for generating
- maps: some code for forward kinematics
- simulator: a copy of this repo that we are freezing in time
- simulator_files: Start here! all the code for actually running armtd is contained in this folder.
Make sure that you have added the full arm_planning repo to your MATLAB path, as well as the CORA_2018 toolbox.
We will update this repo, and clean it up, to be more user-friendly. For now, check out the files in this folder:
arm_planning > simulator_files > examples
All the code here was written by Patrick Holmes, Bohao Zhang, Shreyas Kousik, Daphna Raz, and Corina Barbalata.