The Marching Problem for Autonomous Robots

由于自主机器人系统适用于日常和极端情况下的实际和重要任务，许多研究机构都在加紧研究自主机器人系统的项目。在非自治机器人系统中，不存在控制其他机器人的领导机器人。在这样的系统中，为了完成某些任务，每个机器人必须通过观察其他机器人来单独决定自己的行动。我们重点研究了行军问题，这是基本的合作问题之一。直观地说，在自主机器人系统中，协作行为的质量取决于机器人共享的知识。举一个简单的例子，让我们考虑一个机器人试图组成一条线的问题。如果机器人知道一些坐标系，它们可以很容易地通过确定线的一端在原点上，线在x轴上来组成一条线。然而，如果他们不分享知识，问题就会变得异常困难。因此，我们研究了哪些知识可以单独获取，特别是共享一个坐标系。我们研究了一个由两个全向移动机器人搬运梯子的合作问题。针对这一问题，我们开发了一种分布式算法，并通过计算机仿真对其进行了评估。此外，在我们的仿真中，该算法成功地由n个机器人携带一个n个多边形。这个问题可以看作是行军问题的一个简单版本，其中机器人的队形是物理固定的。下一步，扩展上述结果将为我们提供关于行进问题的想法：即使它们在实践中不携带物体，它们也可以通过附加一些考虑“虚拟”物体效果的函数来假装携带物体，即它们可以在没有物体的情况下行进。该算法将应用于理研所开发的物理机器人系统。进行了初步实验。我们测试了由梯子连接的两个机器人的反馈控制性能。虽然处理梯子的模块有力传感器，但反馈控制效果不佳。原因是由于机器人是直接连接的，所需的响应更少

Many research institutes are intensively working on projects of autonomous robot systems because of their applicability to practical and important tasks both in usual and extreme situations. In au- tonomous robot systems, there is no leader robot that controls the other robots. To complete some tasks in such a system, each robot has to decide own action individually by observing the other robots.We focused on the marching problem which is one of the basic cooperation problems. Intuitively speaking, in an autonomous robot system, the quality of cooperative actions depends on the knowledge that the robots share. For a simple example, let us consider a problem in which robots try to form a line. If the robots have knowledge on some coordinate system, they can easily form a line by deciding that one end of the line is at the origin and the line is on the x-axis. However, if they do not share the knowledge, the problem becomes exceedingly harder. Therefore, we studied on knowledge which rob … More ots can acquire individually, especially sharing a coordinate system.We worked on a cooperation problem carrying a ladder by two omnidirectional mobile robots. We developped a distributed algorithm for this problem and evaluated it by computer simulation. Fur- thermore in our simulation, the algorithm succeeds carrying an n-polygon by n robots. This problem can be considered as a simple version of the marching problem, in which the formation of the robots is fixed physically. For the next step, extending the above results will give us ideas on the marching problem : Even if they do not carry an object in practice, they can pretend to be carrying it by ap- pending some functions which consider the effects from a "virtual" object, i.e., they can march without objects.The algorithm will be implemented to the physical robot system RIKEN developped. A preliminary experiment was done. We tested the performance of the feedback control of two robots being connected by a ladder. Although the modules which handles the ladder have force sensors, the feedback control did not work well. The reason is that since the robots were directly connected the required response Less