From Solitary to Collective Behaviours: Decision Making and Cooperation

In this chapter, we present a set of experiments that improve on the work presented in Chapter 10, in which we showed how a solitary robot can categorise the environment in which it is placed on the basis of the integration over time of its perceptual flow. Here, we consider a social scenario, in which communication is likely to increase the robustness of the categorisation (see also Ampatzis et al., 2007; Trianni et al., 2007). Robots are placed in two different environments and, according to the environmental contingencies they experience, they should perform the appropriate individual or collective action. From the observer—i.e., distal—point of view, this is yet another categorisation problem in which the robotic group faces a binary choice between two environment types. However, from the robot—i.e., proximal—point of view, the binary choice is to be performed between two different behavioural states: a solitary behaviour and a collective one. In the definition of the evaluation function, we emphasise the importance of evaluating the robots for their ability to switch between behavioural states (see Section 11.1.3). The obtained results show that a number of different strategies can be evolved to solve the given problem. Among these, we show that those solutions that exploit communication perform better, systematically achieving a consensus in the group and reducing the decision errors.