Abstract

Although heuristic search is one of the most successful approaches toclassical planning, this planning paradigm does not apply straightforwardly toGeneralized Planning (GP). This paper adapts the planning as heuristic searchparadigm to the particularities of GP, and presents the first native heuristicsearch approach to GP. First, the paper defines a program-based solution spacefor GP that is independent of the number of planning instances in a GP problem,and the size of these instances. Second, the paper defines the BFGP algorithmfor GP, that implements a best-first search in our program-based solutionspace, and that is guided by different evaluation and heuristic functions.

 

Quick Read (beta)