Monitoring Goal Driven Autonomy Agent’s Expectations Generated from Durative Effects
Monitoring Goal Driven Autonomy Agent’s Expectations Generated from Durative Effects
复制标题
监控目标驱动的自主代理从持续效应中产生的期望
DOI：
10.1007/978-3-030-82193-7_32
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Munoz-Avila, Hector
中科院分区：
文献类型：
--
作者：
Reifsnyder, Noah;Munoz-Avila, Hector
鲁棒代理的关键能力之一是自我评估，即代理计算自身边界的能力。评估这些边界的一种方法是使用所谓的“期望”：根据计划、计划实现的目标、初始状态、行动模型和最后执行的行动的函数来定义代理行动过程的边界。在本文中，我们从目标推理文献中重新定义了四种形式的期望，但与那些作品不同的是，代理人以持续行为进行推理。我们提出了属性和比较研究，突出了期望之间的权衡。
One of the crucial capabilities for robust agency is self-assessment, namely, the capability of the agent to compute its own boundaries. A method of assessing these boundaries is using so-called expectations: constructs defining the boundaries of an agent’s courses of action as a function of the plan, the goals achieved by that plan, the initial state, the action model and the last action executed. In this paper we redefine four forms of expectations from the goal reasoning literature but, unlike those works, the agent reasons with durative actions. We present properties and a comparative study highlighting the trade offs between the expectations.