Fast recall of reusable fuzzy plans using acyclic directed graph memory

A planning capability is one of the important features that autonomous real-time systems must have. Fuzzybased planning is more appropriate for planing in realtime dynamic environments such as vehicle navigation and patient monitoring. Planning in such environments needs to be performed as quickly as possible. Planning can be made quicker by reusing portions of similar previous plan segments to efficiently derive a new plan. Planning problems, then, include at least two subproblems. First, the problem of efficiently and effectively generating a plan from scratch. Second, the problem of efficiently and effectively retrieving a plan suitable to be reused and then repairing it to fit the new situation. This paper presents a memory structure as well as fuzzy-based similarity metric for efficiently and effectively retrieving plans specified using fuzzy logic linguistic variables. In the paper, an acyclic directed graph (ADG) model is proposed for memory, such that each node represents an intermediate step in the execution of plans that are represented by other nodes connected to it. Examples of similarity measures computations are presented.