We tested code on ubuntu 20.04 with ROS Noetic, but we didn't use specific features excluding Python3, and you can try other versions First, you need to install Ubuntu 20.04 and ROS Noetic. You can find detailed instructions for installation on sites:
https://old-releases.ubuntu.com/releases/20.04/
http://wiki.ros.org/noetic/Installation/Ubuntu
Also we use Pytorch with CPU calculations
pip3 install torch
This is ROS pkg. Because of this, you need to clone the project into the catkin workspace:
cd ~/catkin_ws/src
git clone https://github.com/warenick/DDP_learning.git
cd ~/catkin_ws ; catkin_make ; source ./devel/setup.bash
We have a launch file for starting preconfigured rviz with the demo of DDP.
roslaunch ddp_learning ddp.launch
In that concrete example of scene H4_mix.py we can see two agents with Social-DDP optimizers moving opposite the Linear extrapolation model.
mix.mp4
We have two main runs - run_solo.py for debugging one agent and run_crowd.py for debugging several agents(crowd) in one scene. We have some prepared scenes for debugging different cases that are kept in scripts/configs/ folder.
Prepared scenes include Agents and optimizer descriptions. For shortness and readability, they have default parameters that may be set to Agents and optimizers.
Generally, for development, we use roslaunch ddp_learning ddp.launch that include rviz, some static_tf, map_server for map reading, and move_base for creating inflated costmap. But main run such run_solo.py or run_crowd.py we start by hands in the iterative development process.
Now we have several Agent options for trajectory optimisation - linear, ddp, social-ddp, costmap-ddp, social-costmap-ddp, but costmap cases still in developing.