Modeling Robot Swarms Using Integrals of Birth-Death Processes

Yara Khaluf¹ and Marco Dorigo² (2015)

¹Department of Computer Science, University of Paderborn, Germany
²IRIDIA, Université Libre de Bruxelles, Belgium

Abstract

This supplementary material contains instructions to download the code to run the cleaning task experiment presented in the main paper (section 5) and explains how to compile and run it.

The ARGoS Simulator

The ARGoS simulator can be downloaded here

The code for the cleaning task

The cleaning task code can be downloaded here

Code compilation

Open a shell, go to the directory where you unpacked the tar.bz2 file and type:
$ mkdir build
$ cd build

To produce fast but not debuggable code, type:
$ cmake -DCMAKE_BUILD_TYPE=Release ..

To produce slow but debuggable code, type:
$ cmake -DCMAKE_BUILD_TYPE=Debug ..

Launch the compilation with the command:
$ make

If you find no error, you are ready to go to the next phase.

Set the environment variable ARGOS PLUGIN PATH to the full path in which the build/ directory is located:
$ export ARGOS_PLUGIN_PATH=/path/to/build/

To run a cleaning task experiment, the configuration files should be in the directory "experiments". Therefore, go to that directory and type:
$ argos3 -c cleaning_robots.argos

Note that the experiment stops when all the robots become inactive.

Parameters

The experiment parameters, that the user can set in the experiment configuration file, are:

• Swarm size: the number of robots participating initially in the cleaning task. This parameter should be set in two places in the experiment configuration file. First, as the value assigned to the "quantity" parameter in the "entity" tag for the "foot-bot" tag. Second, as the value assigned to the parameter "init_num_robots" in the "loop_functions" tag.
• The death (deactivation) rate "mu". This parameter can be found in the "birth_death" tag, within the "params" tag.
• The birth (activation) rate "lambda". This parameter can be found in the "birth_death" tag, within the "params" tag.

The experiment output

An experiments generates two output files. Both are located in the "experiments" directory:

• Active_Inactive_Robots.csv: this file stores the number of robots which become inactive over time and the number of inactive robots that are re-activated (repaired) in the maintenance area and that re-join the cleaning task. Each row consists of three data items: (i) the time; (ii) the number of inactive robots at that time; and (iii) the number of re-activated robots at that time.
• SystemPerformance.csv: this file stores the size of the swarm used in the experiment and the area cleaned by the swarm activity time.

A sample video

In the sample video below, we present a swarm of 50 homogeneous foot-bots performing a cleaning task. The white traces indicate the surface cleaned by the robots while moving in the arena. When the robots have maintenance issues (e.g., their battery is low, they need to be repaired, …) they move to the maintenance area, i.e., the grey area on the right side of the arena. This happens with rate mu. Robots in the maintenance area get repaired at rate lambda (lambda < mu) and then re-join the cleaning tasks.
Robots set their LEDS to green while they are cleaning and to white while they are avoiding obstacles (other robots). When in the maintenance area, robots set their LEDs to red.
The speed of the video is eight times faster than the simulation time. In the code available for downloading, we have removed the visualization of the white traces that indicate the surface cleaned as it slows down the simulation dramatically.