ROLLER: A Lookahead Planner Guided by Relational Decision Trees
نویسندگان
چکیده
In this paper we describe the version of the planner ROLLER submitted to the learning track of the International Planning Competition. This version, learns domain dependent general policies with the aim of improving a lookahead strategy for forward search planning. ROLLER performs the policy learning in a two-step classification process with the relational classifier TILDE. At the first step the classifier captures the preferred operator to be applied in the different planning contexts. At the second step the classifier captures the preferred bindings for each operator in the different planning contexts. In this version of ROLLER a planning context is specified by the helpful actions of the current state, the problem goals, the static predicates of the problem and the last action applied.
منابع مشابه
Ensemble-Roller: Planning with Ensemble of Relational Decision Trees
In this paper we describe the ENSEMBLE-ROLLER planner submitted to the Learning Track of the International Planning Competition (IPC). The planner uses ensembles of relational classifiers to generate robust planning policies. As in other applications of machine learning, the idea of the ensembles of classifiers consists of providing accuracy for particular scenarios and diversity to cover a wid...
متن کاملScaling up Heuristic Planning with Relational Decision Trees
Current evaluation functions for heuristic planning are expensive to compute. In numerous planning problems these functions provide good guidance to the solution, so they are worth the expense. However, when evaluation functions are misguiding or when planning problems are large enough, lots of node evaluations must be computed, which severely limits the scalability of heuristic planners. In th...
متن کاملThe ROLLENT Planning and Learning System at the IPC-8 Learning Track
This paper describes the ROLLENT system submitted to the Eight International Planning Competition, Learning Track. ROLLENT combines two machine learning techniques: generation of entanglements and decision tree learning by ROLLER. Entanglements capture causal relationships for a class of problems while ROLLER learns relational decision trees useful to sort the applicable operators at a given st...
متن کاملLearning Relational Decision Trees for Guiding Heuristic Planning
The current evaluation functions for heuristic planning are expensive to compute. In numerous domains these functions give good guidance on the solution, so it worths the computation effort. On the contrary, where this is not true, heuristics planners compute loads of useless node evaluations that make them scale-up poorly. In this paper we present a novel approach for boosting the scalability ...
متن کاملPlanar Batting with Shape and Pose Uncertainty
This paper explores the planning and control of a manipulation task accomplished in conditions of high uncertainty. Statistical techniques, like particle filters, provide a framework for expressing the uncertainty and partial observability of the real world and taking actions to reduce them. We explore a classic manipulation problem of planar batting, but with a new twist of shape, pose and imp...
متن کامل