Interval computation methods and probabilistic methods for planning and plan checking under uncertainty and incomplete information

TL;DRAbstract

The main problem of planning is to find a sequence of actions that an agent must perform to achieve a given objective. An important part of planning is checking whether a given plan achieves the desired objective. Historically, in AI, the planning and plan checking problems were mainly formulated and solved in a deterministic environment, when the initial state is known precisely and when the results of each action in each state is known (and uniquely determined). In this deterministic case, planning is difficult, but plan checking is straightforward. In many real-life situations, however, the agent does not have complete information about the initial state of the system. In the best case, the agent may know the probabilities of the different fluents that describe the state of the system; sometimes even such probabilities are unknown and the agent has no information whatsoever about the initial state of some fluents. Recently, there has been proposal to use ‘sensing’ actions to plan in

Chat with Paper

AI Agents for this Paper