Planning Heuristics for Conjunctive and Disjunctive Goals

Zhu, Lin. Ph.D., Purdue University, December, 2005. Planning Heuristics for Conjunctive and Disjunctive Goals. Major Professor: Robert L. Givan. Automated planning is an area of Artificial Intelligence (AI) that studies the reasoning process of anticipating the expected outcomes of action sequences and deliberately organizing them to achieve specified goals. Heuristic search based planners have been dominantly successful in recent years. Planning goals can typically be recursively decomposed into conjunctive and disjunctive sets of subgoals, with potential interactions between subgoals. We study the problem of building effective heuristics for achieving conjunctive and disjunctive goals from heuristics for individual subgoals. A major difficulty of planning with conjunctive sets of subgoals is that solutions to them may interfere with each other, yet the subgoals must be simultaneously achieved. Conjunctive subgoals are sometimes best solved sequentially or concurrently, depending on the specific domain structure. Previous methods do not generally accurately reflect interactions between subgoals. We consider a straightforward method for building conjunctive heuristics that smoothly trades off between previous common methods. In addition to first explicitly formulating the problem of designing conjunctive heuristics, our major contribution is the discovery that this straightforward method substantially outperforms previously used methods across a wide range of domains. Based on a single positive real parameter k, our heuristic measure sums the individual heuristic values for the subgoal conjuncts, each raised to the k’th power. Varying k allows loose approximation and combination of the previous min, max, and sum approaches, while mitigating some of the weaknesses in those approaches. Our empirical work shows that for many benchmark plan-