Timeline-based planning system for manufacturing applications

Abstract

In recent years, the Embedded Reasoning Area (ERA) has been developing a planning system targeting fast online planning problems in manufacturing. The planner, which is based on general purpose AI planning techniques, has evolved through several iterations and successfully solved applications such as hyper-modular printers, modular packaging machines, material control for LCD manufacturing, and warehouse management. In this paper, we will describe the core techniques underlying the current version of the planner: a combination of timeline-based state representation and action-based planning algorithm. This combination is proven to be flexible and can quickly adapt to new application and at the same time can scale to complex problems.