III - Autonomous and perceptive systems
The implementation of working and functional models of intelligent
systems is a basic topic in in artificial intelligence research. By
building actual systems or algorithms that have to cope with sensory
signal streams in an uncertain environment, a better understanding of
intelligence and cognition can be obtained. The concept of intelligence
has been redefined since computer chess has become less of a challenge.
The reasons are that for survival in a physical and open environment,
an information processing functionality is needed which is not covered
by the search algorithms that have evolved from traditional AI
research. Also, instead of building hand-crafted separate functions of
intelligence, the current goal of research is to define a physical
system with perceptual-motor ability, place such a system in a physical
context and provide adaptation mechanisms for the system to
autonomously develop functions which are required for adequate
performance in the defined environment. The inspiration for the
implementation of such systems comes from biology, biophysics and human
cognitive science. Selective attention mechanisms in biological vision
are an example of a computationally cheap but pragmatically very
effective method in visual processing. Although the goal is to realize
systems with an integrated architecture for perception, cognition and
action, there will be an accent on perception, as driven by the
requirements of a particular behavior or task. Research streams in the
area of script recognition will remain to operate, covering the
research accent on perceptive systems.
back
|