TY - GEN
T1 - Incremental heuristic search for planning with temporally extended goals and uncontrollable events
AU - Botea, Adi
AU - Ciré, André A.
PY - 2009
Y1 - 2009
N2 - Planning with temporally extended goals and uncontrollable events has recently been introduced as a formal model for system reconfiguration problems. An important application is to automatically reconfigure a real-life system in such a way that its subsequent internal evolution is consistent with a temporal goal formula. In this paper we introduce an incremental search algorithm and a search-guidance heuristic, two generic planning enhancements. An initial problem is decomposed into a series of subproblems, providing two main ways of speeding up a search. Firstly, a subproblem focuses on a part of the initial goal. Secondly, a notion of action relevance allows to explore with higher priority actions that are heuristically considered to be more relevant to the subproblem at hand. Even though our techniques are more generally applicable, we restrict our attention to planning with temporally extended goals and uncontrollable events. Our ideas are implemented on top of a successful previous system that performs online learning to better guide planning and to safely avoid potentially expensive searches. In experiments, the system speed performance is further improved by a convincing margin.
AB - Planning with temporally extended goals and uncontrollable events has recently been introduced as a formal model for system reconfiguration problems. An important application is to automatically reconfigure a real-life system in such a way that its subsequent internal evolution is consistent with a temporal goal formula. In this paper we introduce an incremental search algorithm and a search-guidance heuristic, two generic planning enhancements. An initial problem is decomposed into a series of subproblems, providing two main ways of speeding up a search. Firstly, a subproblem focuses on a part of the initial goal. Secondly, a notion of action relevance allows to explore with higher priority actions that are heuristically considered to be more relevant to the subproblem at hand. Even though our techniques are more generally applicable, we restrict our attention to planning with temporally extended goals and uncontrollable events. Our ideas are implemented on top of a successful previous system that performs online learning to better guide planning and to safely avoid potentially expensive searches. In experiments, the system speed performance is further improved by a convincing margin.
UR - http://www.scopus.com/inward/record.url?scp=78751698813&partnerID=8YFLogxK
M3 - Conference contribution
SN - 9781577354260
T3 - IJCAI International Joint Conference on Artificial Intelligence
SP - 1647
EP - 1652
BT - IJCAI-09 - Proceedings of the 21st International Joint Conference on Artificial Intelligence
PB - International Joint Conferences on Artificial Intelligence
T2 - 21st International Joint Conference on Artificial Intelligence, IJCAI 2009
Y2 - 11 July 2009 through 16 July 2009
ER -