Pugliese, Francesco (2013) Evolution of Cooperative Skills in Social Living Robots. [Tesi di dottorato]

Anteprima

Testo pugliese_francesco_25.pdf Download (6MB) | Anteprima

Abstract

Many animal species, including humans, live in groups, which means, in general terms, they “live together”. In general, we can interpret “living together” more specifically, depending on how two or more individuals live in spatial proximity to each others. Some interesting research questions are: What are the adaptive advantages of the social living, beyond the reproductive needs or looking after the neighbors? In other words, why animals exhibit different degrees and types of sociality? Many authors argue that cooperation enables groups of individuals to reach common goals that are precluded to a single, as for example, in social grooming. Moreover laboratory dyadic cooperation has been analyzed with new simplified paradigms such as “The Loose String Task”. On the other hand, living in non-dyadic groups (consisting of more than two subjects) poses a fundamental problem of social coordination: it is a not simple negotiation problem that in the examples. For the ethology, groups of animals are autonomous units which allow members to synchronize some activities, such as collective foraging and coordination in moving. Evolutionary biologists use the term leadership for behaviors that influence the type, timing and duration of group activity and generally they argue that the reason for the emergence of leadership / followership patterns is the need to coordinate. Furthermore, game-theoretical analysis has shown how, in some situations, the emergence of leadership is almost inevitable. Some experiments, conducted on humans, underline, not only how leadership can emerge in human beings as well as animals, but variation in temperaments may represent a prerequisite for the emergence of leadership. These studies suggest the thesis that evolution has fashioned a so large variation of individuals’ personalities to foster the emergence of leader-follower patterns that are, in turn, essential for solving all the social coordination problems. On one hand, our purpose is finding an answer to questions like: May really arise leadership in a group of genetically heterogeneous robots? Who is the leader? What are leaders made of? What are characteristics and skill of leader? We simulate groups of embodied and artificially evolved robots which must cooperate in order to reach a collective purpose. In every experiment, we try to maintain a strong link between “phenomenon” and “task” derived from experiments on animal behavior, in order to get insights from this kind of data reciprocally. On the other hand, we can contribute to build a new generation of autonomous robotics applications or a new generation of software agents which need a coordination and a leadership emergence to work properly. In this work I illustrate 4 different experimental setups, which examine the mentioned problems under different viewpoints. Results show that sociality give the groups many advantage: 1) sociality facilitate the emergence of more probability to find the food in spite of the increased physical obstruction. Moreover, individual physical limits can be compensated by an increase of the population members number. The sociality fosters the intra-species or intra-race cohesion that allows members (belonging to one species or one race group) to be more successful respect to other species or other races groups; 2) in dyadic cooperative subjects, sociality contribute to the coordination of the group via many communication channels (visual or voiced); 3) in non-dyadic cooperative subjects (i.e. in groups of more than 2 member), social coordination causes the spontaneous emergence of flocking behaviors and leadership. Leaders seem to be the most explorative individuals, the fastest to reach the food areas, etc.

Downloads

Actions (login required)

Modifica documento