Issues addressed
The issues addresses by these architectures extend the
Brooksian Issues.
ATLANTIS
- Time available to perform a task is limited
- The world is largely unpredictable
- Sensors and actuators are imperfect
- Want to respond quickly to unexpected events
- Want to be robust in the face of errors and noise
- Want to be flexible and goal-directed
- Can use, but not require, world knowledge at various levels of abstraction
PCA
- Representational knowledge must be efficient, even if based on first-
principle models.
- "Deductive methods must be complemented by inductive capability for
discovering useful object descriptions."
- Sensing is always motivated in service of some goal.
- Sensing and acting are tightly coupled when sensing has limited purvue
of the environment.
- There is a trade-off between model-based and index-based knowledge storage.
ANA
- How to implement higher-level cognitive functions like learning and planning
in behavior-based agents.
TCA
- How to manage limited physical and computational resources through
coordination.
Return to the Top of this architecture.