ITR: Decision-Theoretic Planning with Constraints

  • Goldsmith, Judith (PI)
  • Dekhtyar, Alexander (CoI)
  • Finkel, Raphael (CoI)
  • Goldstein, Beth (CoI)
  • Marek, Victor (CoI)
  • Mazur, Joan (CoI)
  • Truszczynski, Miroslaw (CoI)

Grants and Contracts Details


This grant supports fundamental and applied research on modeling and automating decision making under uncertainty. The project is developing tools for eliciting, modeling, and managing complex and probabilistic information, algorithms for building models based on this information, and planning algorithms. These will all be applied to the Kentucky Welfare to Work program and to academic advising at the University of Kentucky. The algorithms and software developed will be available for use in other domains. Given formal models of, for instance, Welfare to Work's offerings and of the constraints inherent in legal requirements, logistics, and an individual client's goals and preferences, planning algorithms and heuristics find a plan that optimizes the client's expected utility. (The outcome for a given client in a given program is never certain before the program begins.) The project is developing a unified suite of tools for extracting and eliciting probabilistic information and managing it in a flexible and efficient database management system, a formalism and software for eliciting constraints, preferences and goals, and planning algorithms that take into account both hard and soft constraints. The applications will engage clients and service providers in the Welfare to Work system and students at the University of Kentucky in the processes of determining their own goals and preferences and in planning that will determine whether those goals and preferences are satisfiable. The work will help them strive for maximum satisfaction.
Effective start/end date9/1/038/31/09


  • National Science Foundation: $1,287,000.00


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.