Representation and reasoning for autonomous agents

自主代理的表示和推理

• 批准号: 918-2013
• 负责人: Levesque, Hector
• 金额: $ 2.62万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=635073
• 关键词: Representation; reasoning; autonomous; agents

This research concerns the reasoning that a robot would have to do to interact properly with a world that is changing and incompletely known. Almost all the research in robotics today concerns basic level tasks such as dealing with noisy sensors (like sonar), controlling inexact manipulators (like robot arms), or determining how to move safely without colliding into people or obstacles. Higher level decisions, such as when a robot needs to move at all, in what order it should tackle its jobs, or what it should try to do if something goes wrong, usually end up being scripted in advance. The goal of the research proposed here is to investigate what it would take for an automated agent to be able to make such decisions for itself, based on it knowing what it is trying to do, what it knows about its world, and what it has been able to find out using its sensors. To get these decisions to come out of a general reasoning process, however, it will be necessary to express in a language suitable for automated reasoning, enough of what the agent needs to know about its world, goals, actions, ability, the cognitive states of other agents, collaborative task execution, and so on. The proposed research would examine various aspects of these problems, and attempt to develop high level agent controllers of this sort.

这项研究涉及机器人必须进行的推理，才能与一个正在变化且不完全已知的世界进行适当的互动。如今，几乎所有的机器人学研究都涉及基本层面的任务，比如处理有噪音的传感器(如声纳)，控制不精确的机械手(如机械臂)，或确定如何在不撞到人或障碍物的情况下安全移动。更高级别的决定，比如机器人什么时候需要移动，它应该以什么顺序处理工作，或者如果出现问题，它应该尝试做什么，通常最终都是事先编写好脚本。这里提出的这项研究的目标是调查自动化代理需要做些什么才能为自己做出这样的决定，这是基于它知道自己试图做什么，它知道自己所在的世界，以及它能够使用传感器找到什么。然而，要使这些决策来自一般的推理过程，就必须用一种适合于自动推理的语言来表达，足够多的代理需要知道它的世界、目标、动作、能力、其他代理的认知状态、协作任务执行等。拟议的研究将检查这些问题的各个方面，并试图开发这类高级代理控制器。