Automated Process Planning for CNC Machining.

This article describes an application of AI planning to the problem of automated process planning for machining parts, given raw stock and a CAD file describing the desired part geometry. Researchers at PARC have found that existing planners from the AI community fall short on several requirements, most importantly regarding the expressivity of state and action representations and the ability to exploit domain‐specific knowledge to prune the search space. In this article I describe the requirements for this application and what kind of results from the planning community helped most. Overall, in this project as well as others, I found that even significant results from domain‐independent planning may not be relevant in practice.