This paper describes O-Plan, an architecture for a planning system. The O-Plan strategy seems to be intended for development of large planning systems that may include experimental components representing ongoing research. The architecture therefore focuses on the capability for clear communication between diverse components via a set of general planning and control algorithms.

The paper offers a fairly clear discussion of the design of O-Plan, its representations, and algorithms. Of particular interest is the discussion of some real-world issues, such as user interaction with the planner and the possibility of the planner invoking problem solvers (or knowledge sources) that may fail. These issues are treated well, with an interesting extension to the planning architecture paradigm that allows the handling of these problems in a domain-independent fashion. This will be an interesting paper to read if you are embarking upon or currently building systems that do planning in the practical environment.