Abstract

Designing robots has usually implied knowing beforehand the tasks to be carried out and in what domains. However, in the case of fully autonomous robots this is not possible. Autonomous robots need to operate in an open-ended manner, that is, deciding on the most interesting goals to achieve in domains that are not known at design time. This obviously poses a challenge from the point of view of designing the robot control structure. In particular, the main question that arises is how to endow the robot with a designer defined purpose and with means to translate that purpose into operational decisions without any knowledge of what situations the robot will find itself in. In this paper, we provide a formalization of motivation from an engineering perspective that allows for the structured design of purposeful robots. This formalization is based on a definition of the concepts of robot needs and drives, which are related through experience to the appropriate goals in specific domains. To illustrate the process, a motivational system to guide the operation of a real robot is constructed using this approach. A series of experiments carried out over it are discussed providing some insights on the design of purposeful motivated operation.

中文翻译：

---

动机是设计终身学习机器人的工具

摘要

设计机器人通常意味着事先知道要执行的任务以及在哪些领域中。但是，对于全自动机器人，这是不可能的。自主机器人需要以开放式方式进行操作，也就是说，决定在设计时未知的领域中要实现的最有趣的目标。从设计机器人控制结构的角度来看，这显然带来了挑战。特别是，出现的主要问题是如何赋予机器人设计者定义的目的，以及在不知道机器人会处于何种情况的情况下将这种目的转化为操作决策的手段。在本文中，我们提供了一种形式化从工程角度出发，以实现针对性机器人的结构化设计。这种形式化基于对机器人需求和驱动器概念的定义，这些概念通过经验与特定领域的适当目标相关联。为了说明这一过程，使用这种方法构建了一个指导实际机器人操作的激励系统。讨论了在其上进行的一系列实验，从而为有目的的动机操作设计提供了一些见识。