Issues of Perceptual Anchoring in Ubiquitous Robotic Systems

In ubiquitous robotic systems, robots are im- mersed in an environment containing an abundance of sensors, actuators and smart objects. In such a system, a robot can acquire information about an object from many different sources, possibly including the object itself. While this r ichness of information opens a new landscape of opportunities, it also adds the fundamental challenge of how to coordinate and integrate all the different types of information which are available. In this paper, we define a general computational framework to address this problem. Our framework is based on an extension of the concept of single-robot perceptual anchoring to the multi-robot case, and it can be applied to any ubiquitous robotics system. To make the framework more tangible, we apply it to a specific type of ubiquitous robotic system, call ed Ecology of Physically Embedded Intelligent Systems, or PEIS- Ecology. We also describe a sample implementation based on fuzzy logic, and present some illustrative experiments. I. I NTRODUCTION Integrating robots and smart environments is an idea which is gaining popularity in the field of autonomous robotics. The proponents of this idea suggest a vision in which a large number of robotic devices embedded in the environment communicate and cooperate in order to perform tasks. Robotic devices may include traditional mobile robots, but also distributed sensors, simple actuators, and home appliances. This vision has been given different names, including ubiquitous robotic systems (9), network robot systems (15), intelligent spaces (12), sensor-actuator networks (8), an d PEIS-Ecology (18). In this paper, we shall generally refer to this vision as "ubiquitous robotics". One of the tenants of this vision is that, by exploiting the cooperation betwee n many simple pervasive robotic devices, an ubiquitous robotic system can perform tasks that are beyond the capabilities of current stand-alone robots. As an example, consider a robot trying to grasp a milk bottle. In an ubiquitous robotic system, this robot would not need to use its camera to determine the properties of the bottle (shape, weight, etc.) which would be needed to compute the grasping parameters — a task which has proven to be elusive during several decades of robotic research. Instead, the bottle itself, enriched with an IC-tag, can hol d this information and communicate it to the robot. Ubiquitous robotic systems add a new dimension to robot- environment interaction. In a traditional robotic system t he robot's interaction with objects in the environment is phys - ically mediated; that is, the properties of the objects are estimated using sensors, and their state can be modified usin g actuators. In a ubiquitous robotic system, a robot has an additional possibility: it can interact with an object dire ctly, through digital communication. The robot can ask an object for its properties, and it can even ask it to perform an action . While the ability to interact with objects via direct com- munication opens a new landscape of opportunities, it also creates a new scientific challenge: how to coordinate and integrate physical and digital interaction with the same object. Consider for instance the situation shown in Figure 1. A robot has seen a green box, which it has internally labeled