A Method for Planning Given Uncertain and Incomplete Information

TL;DR

This paper introduces U-Plan, a planning system that constructs and merges plans under uncertain and incomplete information using Dempster-Shafer theory, enabling efficient decision-making in uncertain environments.

Contribution

The paper presents U-Plan, a novel planning approach that handles uncertainty with Dempster-Shafer theory and constructs super-plans by merging multiple possible world plans.

Findings

U-Plan constructs plans for the most supported initial states.

It merges plans into a super-plan with knowledge acquisition points.

U-Plan is faster than classical planning methods for multiple possible worlds.

Abstract

This paper describes ongoing research into planning in an uncertain environment. In particular, it introduces U-Plan, a planning system that constructs quantitatively ranked plans given an incomplete description of the state of the world. U-Plan uses a DempsterShafer interval to characterise uncertain and incomplete information about the state of the world. The planner takes as input what is known about the world, and constructs a number of possible initial states with representations at different abstraction levels. A plan is constructed for the initial state with the greatest support, and this plan is tested to see if it will work for other possible initial states. All, part, or none of the existing plans may be used in the generation of the plans for the remaining possible worlds. Planning takes place in an abstraction hierarchy where strategic decisions are made before tactical decisions. A super-plan is then constructed, based on merging the set of plans and the appropriately timed acquisition of essential knowledge, which is used to decide between plan alternatives. U-Plan usually produces a super-plan in less time than a classical planner would take to produce a set of plans, one for each possible world.