ros_multi_agent_planning

Multi-agent scenario with decentralized planning

Overview

This is a package to demonstrate multi-agent scenario with decentralized planning in a grid world. The final result can be seen below,

Installation

Building

Make sure that you have ROS installed and workspace is created

git clone https://github.com/vvrs/ros_multi_agent_planning.git
cd ~/catkin_ws
catkin_make

Note: gtest is written only for Dijkstra algorithm, and can be tested by using catkin_make run_tests_ros_multi_agent_planning.

Usage

• Use 'add_agent.launch' to create a node for a new agent

roslaunch ros_multi_agent_planning add_agent.launch

• Run 'plan_server_node' to get the path for a given agent

rosrun ros_multi_agent_planning plan_server_node

Launch Files

add_agent.launch

Use this launch file to add new agent to the environment. When launching the new agent, make sure to provide robot_id. Otherwise, the new node kills the old node because of the name conflict. Use, x,y,theta to provide initial location.

Arguments

• robot_id : Default `robot_0'
• x : Default 0
• y : Default 0
• theta : Default 0

planner_and_viz.launch

This launches planner server node, and rviz node for visualization of the path.

Nodes

agent_node

Publications:

• /agent_feedback [ros_multi_agent_planning/Position]
• /agent_path [ros_multi_agent_planning/Path]

Subscriptions: None

Services:

• /update_goal

plan_server_node

Publications: None

Subscriptions:

• /agent_feedback [ros_multi_agent_planning/Position]

Services:

• /get_plan

paths2rviz_node

Publications:

• /path [nav_msgs/Path]

Subscriptions:

• /agent_path [ros_multi_agent_planning/Path]

Services: None

Test

Run below commands in different terminal tabs

roslaunch ros_multi_agent_planning add_agent.launch

roslaunch ros_multi_agent_planning add_agent.launch robot_id:=agent_2 x:=5 y:=5

roslaunch ros_multi_agent_planning planner_and_viz.launch

Now, you will see two nodes for agents (agent_1_node and agent_2_node) running and RViz window opens.

Use below command to make a service call to update the goal

rosservice call /update_goal "goal:
  x: 5
  y: 5
  theta: 0.0
id:
  data: agent_1" 

rosservice call /update_goal "goal:
  x: 10
  y: 8
  theta: 0.0
id:
  data: agent_2" 

Once the above command is executed, rviz plots the path corresponding to robot. Currently, rviz is configured to show paths of only two agents, namely robot_0 and robot_1.