U.S.-Brazil Workshop on Intelligent Robotic Agents

开发能够智能行动的物理制品一直是人工智能和机器人领域的一个长期目标。从机器人方面来说，重点一直是开发能够展示导航、操纵、感知和动作技能以及协作工作能力的机械代理。事实证明，低级别控制的结果，包括传感器的不确定性，关于几何和物理属性的有限推理，以及对象回避等技能，都被证明是有趣和有用的，但大多数自主机器人代理仍然缺乏单独或协同执行更高级别智能操作的能力。这在一定程度上是因为他们需要以合理的权宜之计进行，而已知的方法显然非常复杂，因此速度很慢。希望最近在人工智能规划、学习、推理和代理之间的合作方面的结果将有助于弥合高级别推理和低级别物理执行之间的差距。这次研讨会的目的是审查这一“搭桥”领域的现有成果和研究议程，同时还在同一领域的美国和巴西研究人员之间以及美国国家科学基金会与其巴西对应组织CNPq之间建立国际联系，后者将为巴西与会者提供支持。

It has been a long-standing goal of both the Artificial Intelligence and the Robotics communities to develop physical artifacts that can act intelligently . From the Robotics side, the priority has been on the development of mechanical agents that would exhibit navigation, manipulation, perception, and action skills, along with the ability to work cooperatively. The results on low-level control, including sensor uncertainty, limited reasoning about geometric and physical properties, and skills like object avoidance have proved interesting and useful, but most autonomous robotic agents still lack the ability to perform higher level intelligent operations either alone or in concert. This is partly because of their need to proceed with reasonable expedience, while the known methods are demonstrably very complex, and therefore slow. It is hoped that recent results in AI planning, learning, reasoning, and cooperation among agents will help in bridging the gap between high-level reasoning and low-level physical execution. The purpose of this workshop is to examine the available results and research agenda in this `bridging` area, while also establishing international links between American and Brazilian researchers working in the same areas and between NSF and its Brazilian counterpart organization CNPq, which will provide support for the Brazilian participants.