https://www.cs.cmu.edu/~illah/PAPERS/socialroboticssurvey.pdf

A survey of socially interactive robots

From the Publisher:
Cynthia Breazeal here presents her vision of the sociable robot of the future, a synthetic creature and not merely a sophisticated tool. A sociable robot will be able to understand us, to communicate and interact with us, to learn from us and grow with us. It will be socially intelligent in a humanlike way. Eventually sociable robots will assist us in our daily lives, as collaborators and companions. Because the most successful sociable robots will share our social characteristics, the effort to make sociable robots is also a means for exploring human social intelligence and even what it means to be human.
Breazeal defines the key components of social intelligence for these machines and offers a framework and set of design issues for their realization. Much of the book focuses on a nascent sociable robot she designed named Kismet. Breazeal offers a concrete implementation for Kismet, incorporating insights from the scientific study of animals and people, as well as from artistic disciplines such as classical animation. This blending of science, engineering, and art creates a lifelike quality that encourages people to treat Kismet as a social creature rather than just a machine. The book includes a CD-ROM that shows Kismet in action.

Designing Sociable Robots

: Sociable humanoid robots are natural and intuitive for people to communicate with and to teach. The author presents recent advances in building an autonomous humanoid robot, named "Kismet," that can engage humans in expressive social interaction. She outlines a set of design issues and a framework that she has found to be of particular importance for sociable robots. Having a human-in-the-loop places significant social constraints on how the robot aesthetically appears, how its sensors are configured, its quality of movement, and its behavior. Inspired by infant social development, psychology, ethology, and evolutionary perspectives, this work integrates theories and concepts from these diverse viewpoints to enable Kismet to enter into natural and intuitive social interaction with a human caregiver, reminiscent of parent-infant exchanges. Kismet perceives a variety of natural social cues from visual and auditory channels, and delivers social signals to people through gaze direction, facial expressions, body posture, and vocalizations. The author presents the implementation of Kismet's social competencies and evaluates each with respect to the following: (1) the ability of naive subjects to read and interpret the robot's social cues; (2) the robot's ability to perceive and appropriately respond to naturally offered social cues; (3) the robot's ability to elicit interaction scenarios that afford rich learning potential; and (4) how this produces a rich, flexible, dynamic interaction that is physical, affective, and social. Numerous studies with naive human subjects are described that provide the data upon which the author bases her evaluations.

Sociable machines: expressive social exchange between humans and robots

Neuroscientists have long recognized the importance of understanding the underlying principles of information processing by large populations of neurons. Methods for Neural Ensemble Recordings explores methods for using electrophysiological techniques for monitoring the concurrent activity of ensembles of single neurons. Since current methods allow one to simultaneously record the extracellular activity of up to 100-150 neurons for days or even weeks, neural ensemble recordings have been used to address long-standing issues in development, learning, memory, sensorimotor integration, sensory information processing, and neuronal plasticity.This title examines the many possible applications for this revolutionarymethod. Each chapter offers a step-by-step description for the implementation of a particular technique or experimental paradigm employing simultaneous multiple electrode recordings. The techniques described can be used in applications that impact a large group of life scientists, including: drug screening (pharmacology) in both in vitro and in vivo preparations, developmental studies and studies of neuronal plasticity, chronic monitoring of neuronal function in behavioral studies, physiological monitoring of neuronal activity in cell cultures and brain slices, physiological monitoring of neuronal activity in neurons trafected with genetic vectors, and chronic monitoring of physiological changes in populations of neurons during learning of new sensorimotor and cognitive tasks.

Methods for Neural Ensemble Recordings

Imitation can be viewed as a means of enhancing learning in multiagent environments. It augments an agent's ability to learn useful behaviors by making intelligent use of the knowledge implicit in behaviors demonstrated by cooperative teachers or other more experienced agents. We propose and study a formal model of implicit imitation that can accelerate reinforcement learning dramatically in certain cases. Roughly, by observing a mentor, a reinforcement-learning agent can extract information about its own capabilities in, and the relative value of, unvisited parts of the state space. We study two specific instantiations of this model, one in which the learning agent and the mentor have identical abilities, and one designed to deal with agents and mentors with difierent action sets. We illustrate the benefits of implicit imitation by integrating it with prioritized sweeping, and demonstrating improved performance and convergence through observation of single and multiple mentors. Though we make some stringent assumptions regarding observability and possible interactions, we briefly comment on extensions of the model that relax these restricitions.

/pdf/accelerating-reinforcement-learning-through-implicit-uul86pjzxd.pdf

John Demiris

Papers

Behavior-based primitives for articulated control

Perceptuo-Motor Primitives in Imitation