By learning a range of possible times over which the effect of an action can take place, a robot can reason more effectively about causal and contingent relationships in the world. However, learning these time windows in a noisy environment where random events interfere can pose a challenge. We present an algorithm for learning the interval [t1 min, t1max] of possible times during which a response to an action can take place, and implement the model on a physical robot for the domains of visual self-recognition and auditory social-partner recognition. The environment model that we use to justify our error bounds assumes that natural environments generate Poisson distributions of random events at all scales. From this assumption, we derive a lineartime algorithm, which we call Poisson threshold learning, for finding a threshold T that provides an arbitrarily small rate of background events λ(T) if such a threshold exists for the specified error rate. We can then use this rate to calculate an expected number of false positives in our sample data and discard them. We implement the principles of our method using a motion detection module as our input stream in the visual domain, and sampled audio energy in the auditory domain. In this way, we find time windows for self-generated motion, self-generated audio, and verbal social responses. We also present data on the distributions of these events, showing that while our self-generated action had a normal distribution, the social events were better modeled by a Poisson process. Finally, we present several applications for which such simple classifiers could potentially prove useful, such as mirror selfrecognition and learning the meanings of the words “I” and “you.”

Abstract — An embodied language-learning system is presented that can learn the correct deictic meanings for the words “I ” and “you. ” The system uses contextual clues from already understood words and sensory information from its environment to determine the most likely grounding for a new word. The system also serves as a model for the phenomenon of pronoun reversal among congenitally blind children, as the system learns that “you ” is its own name when it is blinded. The system is novel among grounded systems in that it learns language by observing interactions between other agents, rather than from a helpful caregiver, and in that it associates words with social roles rather than reasoning about visual appearance alone. Index Terms — pronouns, functional language learning, deixis, pronoun reversal, humanoid robot, grounded language, blind language acquisition I.

Interaction with a remote team of robots in real time is a difficult human-robot interaction (HRI) problem exacerbated by the complications of unpredictable realworld environments, with solutions often resorting to a larger-than-desirable ratio of operators to robots. We present two innovative interfaces that allow a single operator to interact with a group of remote robots. Using a tabletop computer the user can configure and manipulate groups of robots directly by either using their fingers (touch) or by manipulating a set of physical toys (tangible user interfaces). We recruited participants to partake in a user study that required them to interact with a small group of remote robots in simple tasks, and present our findings as a set of design considerations.

The ability to identify the self in a mirror reflection and the ability to use the word “I ” effectively are commonly seen as major milestones in a human infant’s development of a concept of self. In addition, deictic pronouns such as “I ” and “you ” present a technical challenge to computational methods for grounded word learning, which have commonly associated word definitions with sensory patterns instead of pragmatic roles. Here, a robot learns the usage of the words “I ” and “you ” by observing others playing a game of catch, then correctly uses the terms to refer to its mirror image and to a conversational partner, respectively. Word learning occurs by using already understood words (“got the ball”) to infer the referents of spoken sentences. The properties of those referents, including the conversational roles of “speaker ” and “addressee ” as well as properties unique to each person, are then associated with the unknown words, and the significance of these associations ranked via chi-square tests. After sufficient observation of others using “I ” and “you,” the robot’s own usage is correct without any need for supervised learning. To achieve mirror selfrecognition, the robot uses the timing of the visual feedback that results from its arm’s movement. The part of the image that is labeled as “self ” is then treated as the robot’s location in the image for the purpose of responding to the command, “Say who got the ball.” 1.

Most robots are designed to operate in environments that

The Wizard of Oz experiment method has a long tradition of acceptance and use within the field of human-robot interaction. The community has traditionally downplayed the importance of interaction evaluations run with the inverse model: the human simulated to evaluate robot behavior, or “Oz of Wizard”. We argue that such studies play an important role in the field of human-robot interaction. We differentiate between methodologically rigorous human modeling and placeholder simulations using simplified human models. Guidelines are proposed for when Oz of Wizard results should be considered acceptable. This paper also describes a framework for describing the various permutations of Wizard and Oz states.