A More Informative Heuristic Function for Fast Forward Planning
Publish place: 11th Annual Conference of Computer Society of Iran
Publish Year: 1384
نوع سند: مقاله کنفرانسی
زبان: English
View: 1,515
This Paper With 5 Page And PDF Format Ready To Download
- Certificate
- من نویسنده این مقاله هستم
این Paper در بخشهای موضوعی زیر دسته بندی شده است:
استخراج به نرم افزارهای پژوهشی:
شناسه ملی سند علمی:
ACCSI11_224
تاریخ نمایه سازی: 5 آذر 1390
Abstract:
This paper proposes a new variation of the Fast Forward (FF) planning. This new variation, which we called H-FF, was applied to the FF policies that are used to reach to a relaxed plan. Specifically, FF uses the NOOP first policy, but new policy of H-FF, named HSP policy, uses the hsp_value attribute in order to make a relax plan. FF and H-FF were tested on three domains (i.e., Mprime, Mystery, and Blocksworld-4ops). The experiments were based on the heuristic estimation, planning times and the number of actions in the final plan (i.e. the plan size). Results show that H-FF often finds a superior solution with respect to the planning times and the solution length criteria, but with high values in the heuristic estimation. The results are highlighted in Mprime and Mystery domains. The more efficient answer of H-FF in time and step shows that HSP policy is a more informative heuristic than NOOP first policy. So the FF assumption that the better answer in relaxed plan, leads to a better plan in the real problem is not the case.
Keywords:
Authors
Emad Soroush
B.S Student Sharif University of Technology Computer Engineering Dep.
Seyyed Ali Akramifar
Ph.D. Student Sharif University of Technology Computer Engineering Dep
G.Reza Ghassem-Sani
Assistant Professor Sharif University of Technology Computer Engineering Dep
مراجع و منابع این Paper:
لیست زیر مراجع و منابع استفاده شده در این Paper را نمایش می دهد. این مراجع به صورت کاملا ماشینی و بر اساس هوش مصنوعی استخراج شده اند و لذا ممکن است دارای اشکالاتی باشند که به مرور زمان دقت استخراج این محتوا افزایش می یابد. مراجعی که مقالات مربوط به آنها در سیویلیکا نمایه شده و پیدا شده اند، به خود Paper لینک شده اند :