Lots of Research Opportunities
We can do a lot more than is now done, but good problems remain
Need more formalisms than just AI Planning
Conditional, sensing, loop planning still problematic
Planning only good for specific goals
Behavioral Goals not addressed:
e.g., keep the various treasury account ratios balanced
Negative Goals not addressed:
e.g., reduce our overhead
Does not address process protocol specifications that may be required by contract or law
Monitoring and Control
is weak
SWS Challenge
includes a basic RossettaNet protocol that is
difficult
to achieve with deductive synthesis.
@copy 2007 Charles Petrie - permission to reproduce widely with attribution