N. 9 - 18 mag 2011
International info a cura di Cecilia Migali
A main research topic at Istc-Cnr concerns the realization of state-of-the-art anticipatory mechanisms for action-outcome and stimulus-stimulus prediction. Predictive abilities make robots more adaptive as they become able to forecast future dangers and opportunities, and to act proactively. In addition to that, anticipatory mechanisms provide the building blocks for the realization of more advanced decision-making, attention control, reasoning, goal-directed action planning and understanding algorithms.
The main scientific objective of this research line is to develop theories and technologies for proactive and goal-directed robots, which can go beyond the here-and-now of current perception and ultimately behave as guided by internally generated goals rather than simply responding to external stimuli. To this aim, we have developed numerous mechanisms for prediction and mental simulations, and implemented them using multiple methodologies, which include connectionist networks, Bayesian networks, and distributed (schema-based) controllers that operate under bounded computational and attention resources. We are adopting our schema-based design methodology in domains as diverse as visuomotor control, robot navigation, multimodal communication, action understanding, and joint action.
Understanding the evolution of communication and human language is one of the hardest problems in science (Christiansen and Kirby, 2003). Of significant research and practical interest is the related artificial perspective: How can populations of robots develop forms of communication of varying complexity, analogous to animal and human communication? This represents a new field of research (Steels, 2003; Nolfi & Mirolli, 2010) that can advance our knowledge about how communication skills originate and evolve in natural organisms and how we can develop autonomous artifacts able to cooperate to solve real-life problems.
More specifically, we provided one of the first experimental demonstration of how a stable and reliable communication system can emerge despite the problems caused by the conflict of interest between individuals and the need to develop two interdependent skills that are adaptively neutral in isolation: an ability to produce useful signals and an ability to react to signals appropriately (Cangelosi & Parisi, 1998). We studied the role that different factors (e.g. kin selection and genetic relatedness) have in the emergence of stable communication systems, and we identified a new factor --- the producer bias --- that results as a by-product of the need of the robots to internally categorize their sensory-motor experiences in adaptive ways (Mirolli & Parisi, 2008). We studied how communication systems can progressively complexify during the evolutionary process, how ‘signals' and ‘meaning' originate and vary (De Greef and Nolfi, 2010). Finally, we studied the role of the co-adaptation of behavioral and communication skills and how embodied concepts are grounded in robots' sensory-motor experiences (De Greef and Nolfi, 2010). We demonstrated the possibility to co-evolve behavioral and communicative skills allowing a team of robots to solve problems which require sophisticated cooperative and/or collaborative skills (Baldassarre et. al., 2007; Trianni et al. 2007; Trianni & Nolfi, 2009, Sperati, Trianni and Nolfi, 2010).
Per saperne di più: - www.springerlink.com/content/1612-4782/