This thesis addresses some fundamental issues toward the realization of "societies" of robots. This objective requires dealing with large numbers of heterogenous autonomous systems, differing in their bodies, sensing and intelligence, that are made to coexist, communicate, learn and classify, and compete fairly, while achieving their individual goals. First, as in human or animal societies, robots must be able to perform cooperative "behaviors" that involve coordination of their actions, based on their own goals, proprioceptive sensing, and information they can receive from other neighboring robots. An effective way to successfully achieve cooperation is obtained by requiring that robots share a set of decentralized motion "rules" involving only locally available data. A first contribution of the thesis consists in showing how these behaviors can be nicely described by a suitable hybrid formalism, including the heterogenous dynamics of every robots and the above mentioned rules that are based on events. A second contribution deals with the problem of classifying a set of robotic agents, based on their dynamics or the interaction protocols they obeys, as belonging to different "species". Various procedures are proposed allowing the construction of a distributed classification system, based on a decentralized identification mechanism, by which every agent classifies its neighbors using only locally available information. By using this mechanism, members of the society can reach a consensus on the environment and on the integrity of the other neighboring robots, so as to improve the overall security of the society. This objective involves the study of convergence of information that is not represented by real numbers, as often in the literature, rather by sets. The dynamics of the evolution of information across a number of robots is described by set-valued iterative maps. While the study of convergence of set-valued iterative maps is highly complex in general, this thesis focuses on Boolean maps, which are comprised of arbitrary combinations of unions, intersections, and complements of sets. Through the development of an industrial robotic society, it is finally shown how the proposed technique applies to a real and commercially relevant case-study. This society sets the basis for a full-fledged factory of the future, where the different and heterogeneous agents operate and interact using a blend of autonomous skills, social rules, and central coordination.
Behavior Classification, Security, and Consensus in Societies of Robots
2012
Abstract
This thesis addresses some fundamental issues toward the realization of "societies" of robots. This objective requires dealing with large numbers of heterogenous autonomous systems, differing in their bodies, sensing and intelligence, that are made to coexist, communicate, learn and classify, and compete fairly, while achieving their individual goals. First, as in human or animal societies, robots must be able to perform cooperative "behaviors" that involve coordination of their actions, based on their own goals, proprioceptive sensing, and information they can receive from other neighboring robots. An effective way to successfully achieve cooperation is obtained by requiring that robots share a set of decentralized motion "rules" involving only locally available data. A first contribution of the thesis consists in showing how these behaviors can be nicely described by a suitable hybrid formalism, including the heterogenous dynamics of every robots and the above mentioned rules that are based on events. A second contribution deals with the problem of classifying a set of robotic agents, based on their dynamics or the interaction protocols they obeys, as belonging to different "species". Various procedures are proposed allowing the construction of a distributed classification system, based on a decentralized identification mechanism, by which every agent classifies its neighbors using only locally available information. By using this mechanism, members of the society can reach a consensus on the environment and on the integrity of the other neighboring robots, so as to improve the overall security of the society. This objective involves the study of convergence of information that is not represented by real numbers, as often in the literature, rather by sets. The dynamics of the evolution of information across a number of robots is described by set-valued iterative maps. While the study of convergence of set-valued iterative maps is highly complex in general, this thesis focuses on Boolean maps, which are comprised of arbitrary combinations of unions, intersections, and complements of sets. Through the development of an industrial robotic society, it is finally shown how the proposed technique applies to a real and commercially relevant case-study. This society sets the basis for a full-fledged factory of the future, where the different and heterogeneous agents operate and interact using a blend of autonomous skills, social rules, and central coordination.File | Dimensione | Formato | |
---|---|---|---|
PhDThesis_final.pdf
accesso aperto
Tipologia:
Altro materiale allegato
Dimensione
29.14 MB
Formato
Adobe PDF
|
29.14 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/145759
URN:NBN:IT:UNIPI-145759