Quasi-Bayesian Strategies for Efficient Plan Generation: Application to the Planning to Observe Problem

Quasi-Bayesian theory uses convex sets of probability distributions and expected loss to represent preferences about plans. The the­ ory focuses on decision robustness, i .e., the extent to which plans are affected by devi­ ations in subjective assessments of probabil­ ity. Generating a plan means enumerating the actions to be taken and providing infor­ mation about the robustness of the actions. The present work presents plan generation problems that can be solved faster in the Quasi-Bayesian framework than within usual Bayesian theory. We investigate this on the planning to observe problem, i.e., an agent must decide whether to take new observa­ tions or not. The fundamental question is: How, and how much, to search for a "best" plan, based on the precision of probability assessments? Plan generation algorithms are derived in the context of material classifica­ tion with an acoustic robotic probe. A pack­ age that constructs Quasi-Bayesian plans is available through anonymous ftp.