Kleywords: ubiquitous computing,
contextual cognition, semantics, probabilistic robotics.
Compared to classical approaches in which robotic systems are
programmed for a limited number of activities, systems that rely on the
concepts of Ubiquitous Computing can independently predict behaviours
in specific situations.
In this thesis a cognitive model for the control of networked robots,
based on the contextual perception of the environment is developed.
Here, concepts of Ubiquitous Computing that are achieved by developing
an appropriate ontology associated with some decision-making mechanisms
are used. By developing the ontology for a system of robots, a
descriptive model of knowledge to be used in industrial robotic
assembly applications is defined. Decision-making mechanisms are based
on Descriptive Logic achieved within the ontology and on a Bayesian
Network, allowing a sufficient level of abstraction required for making
unambiguous decisions appropriate to the current context.
|
