An Interactive Approach for Satisfying Process Plan Generation

Most existing approaches to computer aided process planning aim at full automation by searching for a plan that is optimal with respect to a pre-specified objective function. Such approaches are often infeasible in practice for three reasons: (i) the search space of potential plans is very large, (ii) optimality metrics are often context sensitive and can only be elicited through user interaction, (iii) because of the importance of process planning, organizations are more interested in process planning assistants that support human expert rather than autonomous planners. In this paper, we will describe a new approach to handle the combinatorics of the search space of process plans and generate a satisfying process plan through interaction with the user. In our planning system, a satisfying plan is either a plan which minimizes the penalty cost associated with evaluation criteria violations, or one that satisfies the expert user. Knowledge about process plans is obtained from the Arizona State University Features Test Bed(ASUFTB), a comprehensive and systematic framework for recognizing and reasoning with features of machinable parts. Our approach can be seen as searching the space of interpretations for a design part as plans set up by ASUFTB. We will discuss the detailed algorithms and experimental results for satisfying process plan generation.