This work presents a new class of algorithms that extend the domain of Navigation Among Movable Obstacles (NAMO) to unknown environments. Efficient real-time algorithms for solving NAMO problems even when no initial environment information is available to the robot are presented and validated. The algorithms yield optimal solutions and are evaluated for real-time performance on a series of simulated domains with more than 70 obstacles. In contrast to previous NAMO algorithms that required a pre-specified environment model, this work considers the realistic domain where the robot is limited by its sensor range. It must navigate to a goal position in an environment of static and movable objects. The robot can move objects if the goal cannot be reached or if moving the object significantly shortens the path. The robot gains information about the world by bringing distant objects into its sensor range. The first practical planner for this exponentially complex domain is presented. The planner reduces the search-space through a collection of techniques, such as upper bound calculations and the maintenance of sorted lists with underestimates. Further, the algorithm is only considering manipulation actions if these actions are creating a new opening in the environment. In the addition to the evaluation of the planner itself is each of this techniques also validated independently.

Thesis (Ph.D)--Electrical and Computer Engineering, Georgia Institute of Technology, 2010. / Committee Chair: Dr. Ayanna M. Howard; Committee Member: Dr. Charles Kemp; Committee Member: Dr. Magnus Egerstedt; Committee Member: Dr. Patricio Vela. Part of the SMARTech Electronic Thesis and Dissertation Collection.

The applications of robotics are changing. Just as computers evolved from the realm of research and extreme novelty tools to now becoming essential components of modern life, robotics is also making a similar transition. With the changes in applications come changes in the user base of robotics. These users will span a broad range of society, but there are some key properties that can be used to characterize them. First they, more often than not, will not be the designers of the robots. Second, they will not have robot control as their primary task while operating the robot. Third, they will not have the resources or the desire to provide all the training that the robot will require, yet they will have the need to fine tune robot performance to their specific needs. Fourth, they will want to use multiple modes of interaction to make the robot accomplish the primary task. Fifth, they will expect and demand that the robot remain safe at all times (safe to humans, pets, or personal property) and expect the robot to be a readily replaceable appliance (cheap). Sixth, they will expect that the robot will be intelligent, at least in the confines of the task at hand. These are some of the key properties that will exist for the new user base. To address some of the needs that will arise because of these properties, we propose work that enables behavior transfer from teacher to robotic student that is facilitated through observation and interaction. Many users in the projected user base will not have exposure to the technologies that enable robotic operation. These users will however have some degree of understanding of how they would like the robot to provide assistance in accomplishing the task. The goal of this work is specifically to enable the user to transfer this understanding to the robot, and have the robot acquire this understanding via interactive learning. To make interactive learning possible via interaction we believe that the robot will have to be able to perform some degree of self regulation. Further, since it is assumed that the user will not have access to the robot's internal machinations, the robot will also have to be able to properly manage the knowledge it acquires over time and to verify and validate its understanding periodically. Scaffolding, a method in which teachers provide support while the student learns to master portions of a task, is likely to be the primary method to facilitate this process. This research will undertake study of coherence and its relevance to learning by observation. It will also implement the components that would enable a robot to learn to perform a small set of tasks and demonstrate them in various settings. For this work a robot will be defined as a hardware platform upon which a software agent operates. It is our desire that this software agent will be equipped to operate on any platform and learn any task that a human could perform with the same resources.

The recent advancement of robotic technology brings robots closer to assisting us in our everyday spaces, providing support for healthcare, cleaning, entertaining and other tasks. In this dissertation, I refer to these robots as everyday robots. Scholars argue that the key to successful human acceptance lies in the design of robots that have the ability to blend into everyday activities. A challenge remains; robots are an autonomous technology that triggers multi-faceted interactions: physical, intellectual, social and emotional, making their presence visible and even obtrusive. These challenges need more than technological advances to be resolved; more human-centered approaches are required in the design. However to date, little is known about how to support that human-centered design of everyday robots. In this thesis, I address this gap by introducing an initial set of design guidelines for everyday robots. These guidelines are based on four empirical studies undertaken to identify how people live with robots in the home. These studies mine insights about what interaction attributes of everyday robots elicit positive or negative user responses. The guidelines were deployed in the development of one type of everyday robot: a senior-care robot called HomeMate. It shows that the guidelines become useful during the early development process by helping designers and robot engineers to focus on how social and emotional values of end-users influence the design of the technical functions required. Overall, this thesis addresses a question how we can support the design of everyday robots to become more accepted by users. I respond to this question by proposing a set of design guidelines that account for lived experiences of robots in the home, which ultimately can improve the adoption and use of everyday robots.

The vision of personal robot assistants continues to become more realistic with technological advances in robotics. The increase in the capabilities of robots, presents boundless opportunities for them to perform useful tasks for humans. However, it is not feasible for engineers to program robots for all possible uses. Instead, we envision general-purpose robots that can be programmed by their end-users. Learning from Demonstration (LfD), is an approach that allows users to program new capabilities on a robot by demonstrating what is required from the robot. Although LfD has become an established area of Robotics, many challenges remain in making it effective and intuitive for naive users. This thesis contributes to addressing these challenges in several ways. First, the problems that occur in teaching-learning interactions between humans and robots are characterized through human-subject experiments in three different domains. To address these problems, two mechanisms for guiding human teachers in their interactions are developed: embodied queries and teaching heuristics. Embodied queries, inspired from Active Learning queries, are questions asked by the robot so as to steer the teacher towards providing more informative demonstrations. They leverage the robot's embodiment to physically manipulate the environment and to communicate the question. Two technical contributions are made in developing embodied queries. The first is Active Keyframe-based LfD -- a framework for learning human-segmented skills in continuous action spaces and producing four different types of embodied queries to improve learned skills. The second is Intermittently-Active Learning in which a learner makes queries selectively, so as to create balanced interactions with the benefits of fully-active learning. Empirical findings from five experiments with human subjects are presented. These identify interaction-related issues in generating embodied queries, characterize human question asking, and evaluate implementations of Intermittently-Active Learning and Active Keyframe-based LfD on the humanoid robot Simon. The second mechanism, teaching heuristics, is a set of instructions given to human teachers in order to elicit more informative demonstrations from them. Such instructions are devised based on an understanding of what constitutes an optimal teacher for a given learner, with techniques grounded in Algorithmic Teaching. The utility of teaching heuristics is empirically demonstrated through six human-subject experiments, that involve teaching different concepts or tasks to a virtual agent, or teaching skills to Simon. With a diverse set of human subject experiments, this thesis demonstrates the necessity for guiding humans in teaching interactions with robots, and verifies the utility of two proposed mechanisms in improving sample efficiency and final performance, while enhancing the user interaction.

Thesis (Ph.D.)--University of Delaware, 2009. / Principal faculty advisor: James C. Galloway, Dept. of Physical Therapy. Includes bibliographical references.