Antoine Ligot, Ken Hasselmann, and Mauro Birattari (January 2019)
|
Table of Contents |
We present Arlequin, an off-line automatic design method that produces control software for robot swarms by combining behavioral neural-network modules generated via neuro-evolution. The neural-network modules are automatically generated once, in a mission-agnostic way. With Arlequin, our goal is to reduce the human intervention that is required for the implementation or the operation of previously published modular design methods. Simultaneously, we assess whether neuro-evolution can be used in a modular design method to produce control software that crosses the reality gap satisfactorily. We present robot experiments in which we compare Arlequin with Chocolate, a state of the art modular design method, and EvoStick, a traditional neuro-evolutionary swarm robotics method. The preliminary results suggest that automatically combining neural-network modules into probabilistic finite state machines is a promising approach to the automatic conception of control software for robot swarms.
The instance of control software generated by Arlequin, Chocolate, and EvoStick, as well as their performances are available for download: ControlSoftware.zip and Results.csv
The source code used to generate these instances of control software are available here: Arlequin, Chocolate and EvoStick.
To execute the design methods, you will first need to install the following ARGoS3 libraries: ARGoS3-epuck and epuck-dao.
The robots must aggregate on one of the two black areas present in the arena.
The robots must retrieve as many objects from source areas and deposit them in a nest. The source areas are represented by black circles, the nest is represented by a white area, and the objects are virtual. Indeed, in this version of foraging, a robot is deemed to carry an object after it entered one of the source areas and to retrieve the object when it then enters the nest. A light source is placed behind the nest.