Learning search control rules for plan-space planners: factors affecting the performance

Given the intractability of domain-independent planning, learning effective search control knowledge is vitally important. One way of learning search control knowledge is to use explanation based learning (EBL) methods. This paper aims to analyze and understand the factors influencing the effectiveness of EBL. We present an EBL framework for UCPOP, a partial order planner, and use it to systematically analyze the effect of (i) expressive action representations (ii) domain specific failure theories and (iii) sophisticated backtracking strategies on the utility of EBL. Through empirical studies, we demonstrate that expressive action representations allow for more explicit domain representations which in turn increase the ability of EBL to learn from analytical failures, and obviate the need for domain specific failure theories. We also explore the strong affinity between dependency directed backtracking and EBL in planning.