Cooperative social robots: accompanying, guiding and interacting with people

• Autores: Anaís Garrell Zulueta
• Directores de la Tesis: Alberto Sanfeliu Cortés (dir. tes.)
• Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2013
• Idioma: inglés
• Tribunal Calificador de la Tesis: Yiannis Demiris (presid.), René Alquézar Mancho (secret.), Miguel Ángel Salichs Sánchez-Caballero (voc.)
• Materias:
  • Matemáticas
    • Análisis numérico
      • Construcción de algoritmos
    • Investigación operativa
      • Cibernética
• Enlaces
  • Tesis en acceso abierto en:  TDX  Digital CSIC 
• Resumen

  • The development of social robots capable of interacting with humans is one of the principal challenges in the field of robotics. More and more, robots are appearing in dynamic environments, like pedestrian walkways, universities, and hospitals; for this reason, their interaction with people must be conducted in a natural, gradual, and cordial manner, given that their function could be aid, or assist people. Therefore, navigation and interaction among humans in these environments are key skills that future generations of robots will require to have. Additionally, robots must also be able to cooperate with each other, if necessary. This dissertation examines these various challenges and describes the development of a set of techniques that allow robots to interact naturally with people in their environments, as they guide or accompany humans in urban zones. In this sense, the robots' movements are inspired by the persons' actions and gestures, determination of appropriate personal space, and the rules of common social convention. The first issue this thesis tackles is the development of an innovative robot-companion approach based on the newly founded Extended Social-Forces Model. We evaluate how people navigate and we formulate a set of virtual social forces to describe robot's behavior in terms of motion. Moreover, we introduce a robot companion analytical metric to effectively evaluate the system. This assessment is based on the notion of "proxemics" and ensures that the robot's navigation is socially acceptable by the person being accompanied, as well as to other pedestrians in the vicinity. Through a user study, we show that people interpret the robot's behavior according to human social norms. In addition, a new framework for guiding people in urban areas with a set of cooperative mobile robots is presented. The proposed approach offers several significant advantages, as compared with those outlined in prior studies. Firstly, it allows a group of people to be guided within both open and closed areas; secondly, it uses several cooperative robots; and thirdly, it includes features that enable the robots to keep people from leaving the crowd group, by approaching them in a friendly and safe manner. At the core of our approach, we propose a "Discrete Time Motion" model, which works to represent human and robot motions, to predict people's movements, so as to plan a route and provide the robots with concrete motion instructions. After, this thesis goes one step forward by developing the "Prediction and Anticipation Model". This model enables us to determine the optimal distribution of robots for preventing people from straying from the formation in specific areas of the map, and thus to facilitate the task of the robots. Furthermore, we locally optimize the work performed by robots and people alike, and thereby yielding a more human-friendly motion. Finally, an autonomous mobile robot capable of interacting to acquire human-assisted learning is introduced. First, we present different robot behaviors to approach a person and successfully engage with him/her. On the basis of this insight, we furnish our robot with a simple visual module for detecting human faces in real-time. We observe that people ascribe different personalities to the robot depending on its different behaviors. Once contact is initiated, people are given the opportunity to assist the robot to improve its visual skills. After this assisted learning stage, the robot is able to detect people by using the enhanced visual methods. Both contributions are extensively and rigorously tested in real environments. As a whole, this thesis demonstrates the need for robots that are able to operate acceptably around people; to behave in accordance with social norms while accompanying and guiding them. Furthermore, this work shows that cooperation amongst a group of robots optimizes the performance of the robots and people as well.