Externalizing Internal State

Abstract

Current autonomous robots that are highly reactive are not significantly intelligent and the robots that are significantly intelligent are not highly reactive. The previous research has concentrated on modifications to internal computational structures of robots, ignoring the modifications to external environments (which can preserve both intelligence and reactivity). This work is the first to formalize the modification of an environment that externalizes the internal states. Since some reactive robots exhibited problems like deadlocks and myopic functionality, hybrid architectures with modules like planners began to be explored. In this transition, the potential of reactivity went largely unexamined. Making a robot more reactive means transforming its internal state into an external state that can be extracted through perception (we do not use response time to measure the degree of reactivity, though the response time is important). Some internal states have to be updated whenever external world changes and externalizing these states eliminates such updates, since the most recent information is available in the world itself. Hence there are reasons for robots to be more reactive. In the mobile robot competitions of the American Association of Artificial Intelligence, the rocks in the event of finding life on Mars were painted black to aid in visual recognition. However this kind of environmental modification has not been formally incorporated into the architectures of autonomous robots. As a result, it has been viewed more as a low level fix rather than as a paradigm that deserves a separate investigation. Our work bridges this gap. The states are externalized through environment modifiers called markers. These markers are semantically equivalent to the internal states that they externalize, e.g. a blue strip of paper (a marker) on the door of refrigerator can be used to recognize the refrigerator and thus the presence of the blue strip is equivalent to the presence of the refrigerator (when these strips are put only on the refrigerators). An introduction of markers requires a modification to the behaviors’ stimuli, so that the markers can play an active role in the functionality of a robot. Markers can be kept only on certain ob-