Authors:
Otakar Trunda
and
Roman Barták
Affiliation:
Charles University, Faculty of Mathematics and Physics, Czech Republic
Keyword(s):
Heuristic Learning, Automated Planning, Machine Learning, State Space Search, Knowledge Extraction, Zero-learning, STRIPS, Neural Networks, Loss Functions, Feature Extraction.
Abstract:
Automated planning deals with the problem of finding a sequence of actions leading from a given state to a desired state. The state-of-the-art automated planning techniques exploit informed forward search guided by a heuristic, where the heuristic (under)estimates a distance from a state to a goal state. In this paper, we present a technique to automatically construct an efficient heuristic for a given domain. The proposed approach is based on training a deep neural network using a set of solved planning problems from the domain. We use a novel way of generating features for states which doesn’t depend on usage of existing heuristics. The trained network can be used as a heuristic on any problem from the domain of interest without any limitation on the problem size. Our experiments show that the technique is competitive with popular domain-independent heuristic.