The use of a domain-driven symbolic planner may provide in- teresting performances, even with the most challenging plan- ning domain. However, sometimes a domain utility-function to be maximized does not exist: there are cases in which cre- ating such a function is difficult and error-prone. This paper investigates an alternative approach to afford deterministic planning when no utility-functions are available. In cases like these, classical planning may provide bad performances. The use of a MonteCarlo approach, in conjunction with a goal- based heuristic, has given promising results.
MonteCarlo Tree Search with Goal-Based Heuristic
Luca Sabatucci
2020
Abstract
The use of a domain-driven symbolic planner may provide in- teresting performances, even with the most challenging plan- ning domain. However, sometimes a domain utility-function to be maximized does not exist: there are cases in which cre- ating such a function is difficult and error-prone. This paper investigates an alternative approach to afford deterministic planning when no utility-functions are available. In cases like these, classical planning may provide bad performances. The use of a MonteCarlo approach, in conjunction with a goal- based heuristic, has given promising results.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.