Two Methodologies Toward Artificial Tactile Affordance System in Robotics
Two Methodologies Toward Artificial Tactile Affordance System in Robotics
复制标题
机器人中人工触觉可供系统的两种方法
DOI：
10.21307/ijssis-2017-403
复制
发表时间：
2010
期刊：
影响因子：
--
通讯作者：
H. Yussof
中科院分区：
文献类型：
--
作者：
M. Ohka;Naoto Hoshikawa;Jiro Wada;H. Yussof
摘要如果将负担理论应用于机器人，并不总是需要在其计算机上执行识别和规划的整个过程。由于机器人的触觉感知对于执行任何任务都很重要，所以我们将重点放在触觉感知上，并引入了一个新的概念，称为人工触觉启示系统(ATAS)。它的基本思想是实现一种递归机制，在这种机制中，从对象获得的信息和机器人诱导下一个行为所执行的行为。我们打算基于以下两种方法实现ATAS：(1)将每个规则转换成算法后，基于该算法编写程序模块；ATAS由多个程序模块组成，并根据传感器信息从模块集中选择一个模块；(2)将一组规则表示为由传感器输入列和行为输出列组成的表，表行与规则相对应；由于每个规则被转换为0和1的字符串，因此我们使用遗传算法将由规则串组成的长串视为一个基因，以获得适应其环境的最优基因。对于方法一，我们建立了一个由3到5个模块组成的ATAS，以完成对象抓取、拾取和放置、旋盖和组装等任务。采用方法1，一个配备光学三轴触觉传感器的双手手臂机器人执行上述任务。对于方法2，我们提出了进化行为表系统(EBTS)，它使用遗传算法来获取多个移动机器人的自主合作行为。在验证实验中，三个配备了行为表的代理将对象传递到指定的目标，得分高于四个代理的条件。由于冗余代理不中断其他代理，因此该代理基于其环境信息获得不中断其他代理的集体行为。方法1对于处理类人机器人的任务这样的精细控制非常有效，而方法2对于获得适合环境的一般机器人行为非常有用。
Abstract If the theory of affordance is applied to a robot, performing the whole process of recognition and planning is not always required in its computer. Since the tactile sensing of a robot is important to perform any task, we focus on tactile sensing and introduce a new concept called the artificial tactile affordance system (ATAS). Its basic idea is the implementation of a recurrent mechanism in which information obtained from the object and the behavior performed by the robot's inducing the next behavior. We intend to implement ATAS based on the following two methodologies: (1) after each rule is transformed into an algorithm, a program module is coded based on the algorithm; ATAS is composed of several program modules, and a module is selected from the set of modules based on sensor information; (2) a set of rules is expressed as a table composed of sensor input columns and behavior output columns, and the table rows correspond to rules; since each rule is transformed to a string of 0 and 1, we treat a long string composed of rule strings as a gene to obtain an optimum gene that adapts to its environment using a genetic algorithm (GA). For methodology 1, we established an ATAS composed of 3 to 5 modules to accomplish such tasks as object grasping, pick and place, cap screwing, and assembling. Using methodology 1, a two-hand-arm robot equipped with an optical three- axis tactile sensor performed the above tasks. For methodology 2, we propose the Evolutionary Behavior Table System (EBTS) that uses a GA to acquire the autonomous cooperation behavior of multiple mobile robots. In validation experiments, three agents equipped with behavior tables conveyed an object to a specified goal with higher scores than the four-agent condition. Since the redundant agent does not interrupt the other agents, the agent acquires the collective behavior of not interrupting other agents based on its environment information. Methodology 1 is very effective for such fine control as handling tasks of humanoid robots, and methodology 2 is very useful to obtain general robotic behavior that is suitable for the environment.