Diagnosis and Repair Iteration Planning versus N-step Look Ahead Planning

In this paper we compare the performance of planning algorithms distinguishing and iterating between observation (diagnosis) and repair (action) phases with algorithms extending conventional planning methods with observations using n-step look ahead. Diagnosis and repair iteration planning algorithms are an extension of earlier work of Friedrich and Nejdl, n-step look ahead planning including diagnosis and repair actions has been described recently by Sun and Weld. After describing our new diagnosis and repair iteration planning algorithms in detail, we discuss their theoretical performance and give experimental results. Both theoretical analysis and experimental results show, that our implemented DR-algorithms show a performance compareable to 3 to 4-step look ahead planning, while yielding better plans than both 3 and 4-step look ahead planning.