This report reviews "socially interactive robots": robots for which social human-robot interaction is important. We begin by discussing the context for socially interactive robots, emphasizing the relationship to other research fields and the different forms of "social robots". We then present a taxonomy of design methods and system components used to build socially interactive robots. Following this taxonomy, we survey the current state of the art, categorized by use and application area. Finally, we describe the impact of these these robots on humans and discuss open issues.

This paper presents a new appearance-based place recognition system for topological localization. The method uses a panoramic vision system to sense the environment. Color images are classified in real-time based on nearest-neighbor learning, image histogram matching, and a simple voting scheme. The system has been evaluated with eight cross-sequence tests in four unmodified environments, three indoors and one outdoors. In all eight cases, the system successfully tracked the mobile robot's position. The system correctly classified between 87% and 98% of the input color images. For the remaining images, the system was either momentarily confirsed or uncertain, but never classified an image incorrectly.

Abstract-This paper presents the results of an experiment in human-robot social interaction. Its purpose was to measure the impact of certain features and behaviors on people's willingness to engage in a short interaction with a robot. The behaviors tested were the ability to convey expression with a humanoid face and the ability to indicate attention by turning towards the person that the robot is addressing. We hypothesized that these features were minimal requirements for effective social interaction between a human and a robot. We will discuss the results of the experiment and their implications for the design of socially interactive robots.

PERVASIVEhas one standard for communication between all agents: an agent communication language. ACLs incur overhead that makes them impractical or infeasible for the transfer of low-level data such as video, audio, sensory, or telemetry data. In response to these opposing needs for high-and low-level communication, we developed a two-tiered communication hierarchy, allowing additional, more efficient lines for low-level communication. Integrating the two-tiered communication architecture into the RETSINA MAS significantly improved the system's performance 6 on physical robots for USAR (see Figure 2). Simulation environmentThe scarcity and expense of USARcapable robots has severely restricted USAR robotics research. Field research shows that mobility is only one problem hindering effective use of robots for search and rescue. 3 Testing human perception, situational awareness, and teamwork depends on combining sensed data, human interaction, and automation in experiments. Expense, unreliability, and difficulties in running participants in parallel, especially in multirobot experiments, make physical robotics impractical for the large samples, repeated trials, and varied conditions needed for HRI research. To support HRI, a robotic simulation must accurately render the user interface (particularly, camera video), represent robot automation and behavior, and represent the remote environment that links the operator's awareness with the robot's behaviors. By anchoring the simulation to actual platforms and sensors, we hope to extend our experiments to novel, large, and hazardous environments with control of greater numbers of robots.To meet these requirements, we developed USARsim 7,8 a high-fidelity, extensible simulation of the NIST USAR arenas using the Unreal game engine (see Figure 3). The USAR arenas provide a controlled environment for comparing the effectiveness of different robotic designs, control and mapping algorithms, and team regimes. Each arena (yellow, orange, or red) can contain multiple "victims"-mannequins outfitted with thermal signatures, carbon dioxide emitters, and both noise (for example, screams for help) and motion (for example, waving of hands and fingers) as multimodal clues to the victims' vital signs. The quantitative challenge is discover as many victims as possible quickly and convey sufficient information for human rescuers to navigate the disaster and approach the victims. The arenas pose search tasks with varying difficulty on different dimensions. Challenges to mobility progress from the office-like environment of the yellow arena to the nearly impassable rubble of the red arena. Perceptual difficulties vary, from visually confusing patterns, glass panels, mirrors, and sonar-absorbing padding in the yellow arena to the few perceptual difficulties in the red arena.Our simulated environments include these three arenas and the larger, fixed USAR reference site in an abandoned Nike silo on the NIST Gaithersburg campus. Because the Unreal engine uses stan-

Sage is a robot that has been installed at the Carnegie Museum of Natural History as a full-time autonomous member of the staff. Its goal is to provide educational content to museum visitors in order to augment their museum experience. This paper discusses all aspects of the related research and development. The functional obstacle avoidance system, which departs from the conventional occupancy grid-based approaches, is described. Sage's topological navigation system, using only color vision and odometric information, is also described. Long-term statistics provide a quantitative measure of performance over a nine month trial period. The process by which Sage's educational content and personality were created and evaluated in collaboration with the museum's Divisions of Education and Exhibits is explained. Finally, the ability of Sage to conduct automatic long-term parameter adjustment is presented.