engleski

Computation approach for realisation of context-aware robot

Humans are using memories, guesses and other implicit information stored or collected to reason about most appropriate solutions. Unlike humans, robots do not understand context by default. Compared to conventional approaches where robots are preprogramed to react to a finite number of environmental occur- rences, contextual awareness can enable modeling of humanlike adaptation skills. Computational models pre- sented in this work could be understood as context-to-data interpreters that transform contextual information into data, allowing machines to make context- driven decisions. The basic model contains three main parts. The first part is used to track and collect significant environmental information. The second part represents formal knowledge about the domain of interest. The model also contains a probabilistic component realized by a Bayesian Network. The overall methodology is presented through three separate examples illustrating reasoning based on: (i) phenomenon of social capital, (ii) human bodily awareness and (iii) human emotions.

affective robotics ; context-awareness ; probabilistic reasoning ; knowledge representation

nije evidentirano