multi-robot_task_allocation

Algorithms of a multi-robot task allocation for multi-target tracking are provided, which have been developed in ROS Kinetic on Ubuntu 16.04. This package provides several features for a multi-robot task allocation when tracking multiple targets. A local algorithm plays a key role in this algorithm where it provides nice properties on communication limitation and was referred by Floréen, Patrik, et al. "Local approximability of max-min and min-max linear programs." Theory of Computing Systems 49.4 (2011): 672-697.

Run a demo

Demo is included in max_min_lp_demo. In max_min_lp_demo/data there are several instances that you can test. Local algorithm is implemented in a sequential way, not in a distributed since one machine is used to test the performance of the algorithm.

roslaunch max_min_lp_demo test_max_min_lp.launch

In case you want to visualize an original and layered graphs of instance that you are testing, run the following.

rosrun max_min_lp_visualization max_min_lp_plot.py

Then, the result of the graph will be in the max_min_lp_visualization/log folder. The following will activate the demo.

rostopic pub /robot_status std_msgs/String demo

Two main algorithms

There are two versions of multi-robot multi-target tracking algorithms. The first algorithm is a local algorithm based one, which is explained in the beginning. The second algorithm is a greedy one where each robot greedily track targets without considering other robots. For both a central node exists to simulate sensing and communication graphs. Each robot has its own node for local computation and getting odometry information from the gazebo simulation.

Local algorithm

Run the gazebo simulation first. This will spawn five (turtlebot) robots with thirty stationary and moving targets.

roslaunch max_min_lp_simulation five_robot_twenty_target_case.launch

Then, run the main algorithm.

roslaunch max_min_lp_simulation simulation_server_to_robots.launch

To initiate the algorithm,

rostopic pub /robot_status std_msgs/String run

Greedy algorithm

Again, run the gazebo simulation first.

roslaunch max_min_lp_simulation five_robot_twenty_target_case.launch

Then, run the main algorithm.

roslaunch max_min_lp_simulation greedy_server_to_robots.launch

To initiate the algorithm,

rostopic pub /robot_status std_msgs/String run

Generating and testing motion primitives of a robot

In this package, which is max_min_lp_simulation, turtlebot is used as a mobile platform. This feature allows to generate and test different sets of motion primitives of a robot while changing cmd_vel values with different time intervals. In our setting time interval is defined by the number of publishers applied to. First, run the following command.

roslaunch max_min_lp_simulation single_robot_case.launch

This will run the gazebo simulator with a single robot.

rosrun max_min_lp_simulation motion_primitive_gerator

Now you can feed /robot/status rostopic in order to move the robot. In the following the first command moves the robot x direction and the second one rotates the robot in the counter-clockwise. In the gazebo simulator the positive values for x is the forward direction to the robot while the positive values for z rotates the robot in the counter-clockwise.

rostopic pub /robot_status std_msgs/String x
rostopic pub /robot_status std_msgs/String z