Supplementary material for the paper:

From Solitary to Collective Behaviours:
Decision Making and Cooperation

Vito Trianni, Christos Ampatzis, Anders Lyhne Christensen,
Elio Tuci, Marco Dorigo, and Stefano Nolfi

This page contains support material of the paper "From Solitary to Collective Behaviours: Decision Making and Cooperation" submitted to the 9^th European Conference on Artificial Life (ECAL 2007). In particular, we provide a detailed description of the experimental setup, which was left out due to space limit. We also provide an extensive behavioural analysis of the obtained results, supported by videos of the evolved behaviours. Finally, we describe the behaviours produced by all controllers that make use of communication.

• Experimental Setup
• Behavioural Analysis
• Class U
• Class B
• Class M
• Class C
• Communicative Behaviours
• Controller C₄
• Controller C₁₆
• Controller C₇
• Controller C₁₉

We make use of some videos to support the description of the obtained results. In each video, three s-bots move in the arena, in which the way out is either present of not . A green spot on top of the robot indicates that the robot is signalling, and the emitted tone can be perceived by all robots in the arena. All videos are mpeg and weigh about 4M.

The behaviours produced by the successful controllers were grouped in four classes, named U, B, M and C.

Class U= {C₄,C₆,C₁₄,C₁₇} encompasses the "unsuccessful" controllers, that is, those controllers that only result in partial solutions to the problem. These controllers produce appropriate searching behaviour when s-bots are in state S. However, when s-bots are placed in environment B they fail in systematically aggregating

The class B={C₁,C₅,C₈,C₁₀,C₁₆} encompasses controllers that produce a strategy named "bouncing" after the aggregation behaviour of the s-bots in state C, which search for each other by continuously bouncing off the circular band. In the following, we describe the behaviour produced by controller C₁₀

The class M={C₃,C₇,C₁₁,C₁₃,C₁₉,C₂₀} encompasses controllers that produce a strategy named "meeting", due to the fact that s-bots aggregate by encountering at a meeting point, which is normally close to the centre of the arena. Similarly to class B, controllers of this class do not make use of communication (exception made for C₇ and C₁₉, which are described later) and the decision to switch from state S to state C is performed individually. The main difference with class B controllers resides in the aggregation behaviour, that lets robots leave the band and move in circles close to the centre of the arena, waiting for the other s-bots to reach a similar position. In the following, we present the behaviour produced by controller C₂₀

The last class C={C₂,C₉,C₁₂,C₁₅,C₁₈} is named "cooperative" because it encompasses controllers that produce communicative behaviours exploited for cooperation. Here, the perception of a signal triggers the way out searching behaviour. On the contrary, aggregation is possible only if there is no robot signalling in the arena. The switch from state S to state C is coordinately performed when all robots have spent enough time searching for the way out without finding it. On the contrary, if one robot finds the way out, it emits a signal that allows all other s-bots to remain in state S. In the following, we give a detailed description of the behaviour produced by controllerC₁₈, showing how the use of communication can lead to robustness and efficiency.