This repo contains most of the code for my master's thesis - "Optimized Task Coordination for Heterogenous Multi-Robot Systems", which formulated and implemented a solution for for multiple robot task allocation.

It uses a combination of a deterministic greedy algorithm and metaheuristic genetic algorithm to assign tasks to a team of UGVs in an optimal manner. The tasks considered are "go to goal" and "patrol" tasks. The UGVs considered are turtlebots and perform the tasks in a collision free manner using reciprocal velocity obstacles.

This repo contains 2 ROS packages that are intended to be exectued from a docker container (Dockerfile included, along with build and run scripts).

mrta_main - Task allocation algorithm and task scenario descriptions. Additional information on the mrta_main package is on mrta_main/README.md and describes the file structure / usage of this package.

ugv_action_model - Motion planning, state management and task execution for UGVs. Additional inforamtion on the mrta_main package is on ugv_action_model/README.md. and describes the file structure / usage of this package.

/current_directory/optimized_task_coord/
├── Dockerfile
├── dockerbuild.sh
├── dockerrun.sh
├── src/
│   ├── mrta_main/
│   ├── ugv_action_model/

When running the container using dockerrun.sh, the project directory is volume mounted as /workdir. This directory is treated as a ROS workspace.

/workdir/
├── Dockerfile
├── dockerbuild.sh
├── dockerrun.sh
├── src/
│   ├── mrta_main/
│   ├── ugv_action_model/

This repository is intended to be run using the provided docker container. In order to run it without the container, you will have to install the dependancies as well as change file path variables in the .py scripts. A list of dependancies is not provided, but can be reverse engineered from the provided Dockerfile and requirements.txt

1 - Clone the repo and change directory into the repo:

git clone [email protected]:albud187/optimized_task_coord.git

cd optimized_task_coord

2 - Build the container:

sh dockerbuild.sh

3 - Run the container:

sh dockerrun.sh

4 - While inside the container on project directory /workdir, build the ros pakages:

cd /workdir

catkin_make

source devel/setup.bash

5 - Make all .py files executable:

chmod +x ./*

Unless otherwise stated, it is assumed that the setup is completed and you are in the docker container on /workdir

All instructions here concern running a demo scenario (31T-6R) included in this repo. Instructions to create your own scenario are on mrta_admin/README.md

There are two types of demos that can be excuted from this repo, a demo using rqt and a demo using mrta_main to determine the actions taken by the robots. Both them require two terminals as follows:

1 - Open two terminals and run the docker container. On both terminals execute:

sh dockerrun.sh

2 - On terminal 1 execute:

roslaunch ugv_action_model ugv_6.launch

This will launch a gazebo simulation with 6 turtlebots.

The next steps pertain to the rqt demo.

3a - On terminal 2 execute:

rqt

We will use rqt to publish to the r_n/task topics, where r_n is the namespace of an individual robot and ranges from r_0 to r_5 for this demo.

4a - The topics of r_0/task to r_5/task are of type PoseStamped, and describe tasks as follows: PoseStamped.header.frame_id: type of task. String. Set it to "A" for a goal to goal task. Set it to "D" for a patrol task.

PoseStamped.pose.position.x: x coordinate of task location. Float.

PoseStamped.pose.position.y: y coordinate of task location. Float.

PoseStamped.pose.orientation.x: Length of patrol route. Only applicatible if PoseStamped.header.frame_id = "D"

PoseStamped.pose.orientation.w: Task completion indicator. Set it to 1 (or anything not 0) otherwise the robot will not move.

5a - Publish the task topic(s). The robot(s) corresponding to the task topic(s) will execute their tasks. An example with one robot is shown below with one task published to robot r_0, however, you can publish multiple tasks to multiple robots at the same time, and they will execute their tasks at the same time.

The next step pertains to the mrta_main demo.

3b - On terminal 2 execute:

roslaunch mrta_main main.launch scenario:=31T-6R

This will execute the task allocation algorithm with the example scenario 31T-6R. After the algorithm is finished, the task allocation will be shown below. Then after closing this image, the robots will execute their assigned tasks.

This will execute the optimization algorithms to assign tasks described in the src/scenarios/31T-6R to the 6 turtlebots. This scenario has 31 tasks and 6 robots. The tasks are described in src/scenarios/31T-6R/tasks.csv, the suitability between the robots and tasks are described in src/scenarios/31T-6R/suitabilities.csv. The type of robots and their work capacity are described in src/scenarios/31T-6R/agents.csv. Instructions on creating your own scenario are described in mrta_admin/README.md.