Department of Electrical Engineering and Computer Science
University of Wisconsin Milwaukee.
|
- DRIPS - Decision-Theoretic Refinement Planning System DRIPS is a decision-theoretic planner that efficiently identifies optimal plans based on a user-defined utility function and a set of available probabilistic actions grouped into an abstraction/decomposition hierarchy. |
| Investigators |
|
|
| Background |
|
To solve complex problems, decision analysts often must evaluate
numerous potential plans by constructing decision trees by hand.
The need to perform this process manually limits the space of
possible strategies they can explore explicitly, and provides
the potential to introduce errors into the analysis.
|
| Current Status |
|
The DRIPS decision-theoretic refinement planning system automates
the process of decision tree construction and evaluation. The
system uses abstraction to explore a large space of possible
plans efficiently; it composes individual action descriptions to
create decision trees. The system can reason about discrete and
continuous attributes, as well as about time. It finds the plan
or plans that maximize the expected value of a user-defined
utility function.
DRIPS provides a mechanism to identify optimal plans involving diagnosis and treatment. We have evaluated the application of DRIPS to the management of suspected acute deep venous thrombosis (DVT) of the lower extremities. |
| Software |
|
| Publications |
|