Backprop-MPDM torch: Multi-policy Decision-making framework for building robot path in social environment
We tested code on ubuntu 18.04 with ROS Melodic, but we didn't use specific features and you can try other versions First, you need to install Ubuntu 18.04 and ROS melodic. You can find detailed instructions for installation on sites:
https://releases.ubuntu.com/18.04.4/ http://wiki.ros.org/melodic/Installation/Ubuntu
Then, you need to setup ROS for Python3. You can find useful information in that article: https://medium.com/@beta_b0t/how-to-setup-ros-with-python-3-44a69ca36674
In short, you need to install Python 3 libs for ROS with pip3
sudo pip3 install rospkg catkin_pkg
This is ROS pkg, becouse of this, you need to clone project into catkin workspace:
cd ~/catkin_ws/src
git clone https://github.com/alexpostnikov/Backprop-MPDM_torch.git
All project on Pytho3, but you need to compille custom ROS msgs that we made for that project
cd ~/catkin_ws
catkin_make
source ./devel/setup.bash
After that you can check compilled msgs:
rosmsg list | grep mpdm
That command should return you list of msgs like this:
mpdm/Learning
mpdm/Ped
mpdm/Peds
mpdm/Propagation
(all commands for different terminals)
roscore
rosrun mpdm visualiser.py
rosrun mpdm fake_publicator.py
rosrun mpdm node.py
roslaunch mpdm rviz.launch
after that you shold see visualisation of working mpdm node in rviz like this:
All goals would update automatically, but also you can set new goal for robot by picking "2D Nav Goal" in Rviz