i-am-neet / mapf-sim

Multi-Agent Reinforcement Learning (MARL) on Stage simulator with ROS

System

Name	Version
Ubuntu	20.04
Robot Operating System (ROS)	noetic
Python	3.8
Pip	20.0.2
Nvidia driver	460.91.03
CUDA	10.1
cUDNN	7.6
Gazebo	11.5.1

Setup

Create python environment

$ virtualenv -p python3.6 $HOME/mapf
$ source $HOME/mapf/bin/active

Install requirement package

Simulator:

$ sudo apt install screen python3-catkin-pkg
$ sudo apt install ros-noetic-hector-sensors-description
$ sudo apt install ros-noetic-map-server
$ sudo apt install ros-noetic-move-base

Pytorch:

$ pip install torch==1.7.1+cu101 torchvision==0.8.2+cu101 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

Python:

$ pip install -r requirements.txt

Clone & Build this project

$ cd $HOME
$ mkdir -p mapf_ws/src && cd mapf_ws
$ git clone --recursive https://github.com/i-am-neet/mapf-sim.git src/
$ catkin_make

Startup

$ cd $HOME/mapf_ws && source devel/setup.bash
# Start RL env of simualtor
$ roslaunch robot_gazebo home_area.launch gui:=false # or empty.launch
# Bringup robots
# There are 2 arguments for bringup_all.bash:
# $1:PATH of INIT_POSES.CFG and $2:PATH of INIT_GOALS.CFG
$ source src/gazebo/robot_gazebo/scripts/bringup_all.bash
# Quit by
$ killall screen
# MAPF node, each agent needs one
$ rosrun mapf_pkg mapf.py -h
usage: mapf.py [-h] [--current-robot-num CURRENT_ROBOT_NUM]
               [--robots-num ROBOTS_NUM] [--robot-diameter ROBOT_DIAMETER]
               [--map-resolution MAP_RESOLUTION]

Reinforcement Learning Arguments

optional arguments:
  -h, --help            show this help message and exit
  --current-robot-num CURRENT_ROBOT_NUM
                        The number of current robot. (For env) (default: None)
  --robots-num ROBOTS_NUM
                        The amount of all robots. (For env) (default: None)
  --robot-diameter ROBOT_DIAMETER
                        The diameter of robot (default: 0.25, according to
                        TB3) (For env) (default: 0.25)
  --map-resolution MAP_RESOLUTION
                        The resolution of map (default: 0.01) (For env)
                        (default: 0.01)