E-PDDL定义认知规划问题的标准化方法_E-PDDL A Standardized Way of Defining Epist
版权申诉
71 浏览量
2022-01-17
18:14:55
上传
评论
收藏 182KB PDF 举报
arXiv:2107.08739v1 [cs.AI] 19 Jul 2021
E-PDDL: A Standardized Way of Defining Epistemic Planning Problems
Francesco Fabiano
DMIF Departmen t
University of Udine, Italy
francesco.fabiano@uniud.it
Biplav Srivastava
AI Institute
University of South Carolina, Columbia, USA
biplav.s@sc.edu
Marianna Bergamaschi Ganapini
Philosophy Department
Union College, NY, USA
bergamam@union.edu
Jonathan Lenchner and Lior Horesh and Francesca Rossi
Thomas J. Watson Research Center, Yorktown Heights, NY USA
{lenchner; lhoresh}@us.ibm.com
francesca.rossi2@ibm.com
Abstract
Epistemic Planning ( EP) refers to an automated plan-
ning setting where the agent reasons in the space of
knowledge states and tries to find a plan to reach a de-
sirable state from the current state. Its general form,
the Multi-agent Epistemic Planning (MEP) problem in-
volves multiple agents who need to reason about both
the state of the world and the information flow between
agents. In a MEP problem, multiple approaches have
been developed recently with varying restrictions, such
as considering only the concept of knowledge while not
allowing the idea of belief, or not allowing for “com-
plex” modal operators such as those needed to handle
dynamic common knowledge. While the diversity of ap-
proaches has led to a deeper understanding of the prob-
lem space, the lack of a standardized way to specify
MEP problems independently of solution approaches
has created difficulties in comparing performance of
planners, identifying promising techniques, exploring
new strategies like ensemble methods, and making it
easy for new researchers to contribute to this r esearch
area. To address the situation, we propose a unified way
of specifying EP problems - the Epistemic Planning Do-
main Definition Language, E-PDDL. We show that E-
PPDL can be supported by leading MEP planners and
provide corresponding parser code that translates EP
problems specified in E-PDDL into (M)EP problems
that can be handled by several planners. This work is
also useful in building more general epistemic planning
environments where we envision a meta-cognitive mod-
ule that takes a planning problem in E-PDDL, identifies
and assesses some of its features, and autonomously de-
cides which planner is the best one to solve it.
Motivation
Multi-agent scenarios are ubiqu itous in everyda y life. We
often ne ed to make decisions ba sed on informa tion on the
problem to be so lved and on our k nowledge, or belief, about
the preferences and actions of oth er agents involved in, or
impacted by, the decision. It is therefore an epistemic rea-
soning task to make decisions in this scenarios.
Often such decision involve creating a plan, that is, a se-
quence of actions that, when executed, will le a d to a desired
Copyright © 2021, Association for the Advancement of Artificial
Intelligence (www.aaai.org). All rights reserved.
goal state. The research ar e a of Multi-agent Epistemic Plan-
ning (MEP) attempts to build artificial agents that can build
such plans. MEP is a generalization of the concept of E pis-
temic Planning (EP), which refers to a single agent setting
that reasons about knowledge and beliefs. In the MEP sce-
nario, agents also reason about the information that o ther
agents have of each oth e r, and how such informa tion flows
between agents.
Several approa c hes to develop epistemic solver-
s/planners have been made across the community,
see for examp le ( Baral et al. 2015; Muise et al. 2015;
Fabiano e t al. 2020). While these solvers may have different
restrictions, e.g., considering only the conc ept of knowledge
while not allowing the idea of belief, or not allowing for
dynamic common knowledge, all of them share the same
goal: to plan while considering the information flows
between the agents.
While the d iversification of appro aches in solving this
problem certainly brings benefits to the entire community,
with many spec ialized efficient planners that are suitable for
specific scenarios, the lack of a standardized way to define
input problems in multi-agent epistemic planning has led to
the creation of ad-hoc languages th at can become an ob stacle
when comparing, combining (i.e., ensemble), or testing dif-
ferent planners. We believe that a unified way of expressing
MEP problems would provide researchers with a faster a nd
less err or-prone way of defining standardized be nchmarks
and would also relieve them from the burden of having to
learn a new langu a ge each time they nee d to make use of a
different tool. Moreover, it would allow for easy tran sf e r of
techniques a nd results betwe e n different plannin g scenarios
and planner methodologies.
In this p aper we th e refore pro pose the Epistemic Plan-
ning Domain De finition Language (E-PDDL), a unified way
of defining epistemic problems that, similarly to the well-
known language PDDL for classical planning, could be re-
garded as the standard language to be used to express MEP
problems. Being able to define MEP problems in E-PPDL
and then to map them into sp e cific languages for existing
planners provides also the necessary environment for a real-
time ch oice of the best p la nner for a given MEP problem.
Indeed , this work is part of a larger effort that aims to equip
artificial agents with meta-cognitive capabilities that, given
internal values, resour ces, and criteria, can identify some of
资源评论
易小侠
- 粉丝: 6509
- 资源: 9万+
最新资源
- 建筑结构\施工图\B型施工图-建筑-别墅结构施工图.dwg
- 基于stm32的智能家居
- 高分项目，基于Unity3D开发实现的俯视第三人称射击完整项目Unity游戏，内含完整源码+资源
- 强大且美观的WPF UI控件库：HandyControl
- 基于Java开发的景区APP，包含智能推送、景区信息、出行订票、出行交通、出行天气、旅行翻译、客流量实时监测、人数预测和智能分析
- 基于STM32微控制器的简单超声波液位控制器
- 建筑结构\施工图\B型施工图-建筑-平面图.dwg
- 高分项目，基于Unity3D开发实现的俄罗斯方块游戏UnityTetris，内含完整源码+资源
- 建筑结构\施工图\B型施工图-建筑-立面.dwg
- 基于STM32 MCU的Lil32系统控制器
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
