This paper describes the development of algorithms that decide when to move, where to move, and how to look for people in a home environment. We introduce a design framework as a tool to guide the development of a social robot to proactively be with people for companionship and assistance in the home. Through a series of experiments ranging from simulations to longitudinal A/B studies, we demonstrate how to utilize the design framework to help guide the evaluation and selection of solutions. We deployed our autonomous robot in a long-term in-situ study and found our proposed approach to be more capable of being co-present with its household members compared to a baseline approach. Conducted in an industry setting, our research approach departs from typical academic practices as the motivations are inherently different. We share our perspective on the differences of industry research when developing a social robot as a commercial product.

We performed a study that examined the effects of a humanoid robot giving the minimum required feedback - graded cueing - during a one-on-one imitation game played children with autism spectrum disorders (ASD). 12 high-functioning participants with ASD, ages 7 to 10, each played “Copy-Cat” with a Nao robot 5 times over the span of 2.5 weeks. While the graded cueing model was not exercised in its fullest, using graded cueing-style feedback resulted in a nondecreasing trend in imitative accuracy when compared to a non-adaptive condition, where participants always received the same, most descriptive feedback whenever they made a mistake. These trends show promise for future work with robots encouraging autonomy in special needs populations.

We present a user study and dataset designed and collected to analyze how humans use space in face-to-face interactions. In a proof-of-concept investigation into human spatial dynamics, a Hidden Markov Model (HMM) was trained over a subset of features to recognize each of three interaction cues - initiation, acceptance, and termination - in both dyadic and triadic scenarios; these cues are useful in predicting transitions into, during, and out of multi-party social encounters. It is shown that the HMM approach performed twice as well as a weighted random classifier, supporting the feasibility of recognizing and predicting social behavior based on spatial features.

Partially Observable Markov Decision Processes (POMDPs) offer a powerful mathematical framework for making optimal action choices in noisy and/or uncertain environments, in particular, allowing us to merge localization and decision-making for mobile robots. While advancements in POMDP techniques have allowed the use of much larger models, POMDPs for robot navigation are still limited by large state space requirements for even small maps. In this work, we propose a method to automatically generate a POMDP representation of an environment. By using variable resolution decomposition techniques, we can take advantage of characteristics of the environment to minimize the number of states required, while maintaining the level of detail required to find a robust and efficient policy. This is accomplished by automatically adjusting the level of detail required for planning at a given region, with few states representing large open areas, and many smaller states near objects. We validate this algorithm in POMDP simulations, a robot simulator as well as an autonomous robot.

Personal robots are becoming increasingly prevalent, which raises a number of interesting issues regarding the design and customization of interfaces to such platforms. The particular problem addressed by this paper is the use of learning methods to improve the quality and effectiveness of human-machine interaction onboard a robotic wheelchair. In support of this, we present a method for learning and adapting probabilistic models with the aid of a human operator. We use a Bayesian reinforcement learning framework, that allows us to mix learning and execution, as well as take advantage of prior information about the world. We address the problems of learning, handling a partially observable environment, and limiting the number of action requests. We demonstrate empirical feasibility of our approach on an interface for an autonomous wheelchair.

Gesture recognition is becoming a more common interaction tool in the fields of ubiquitous and wearable computing. Designing a system to perform gesture recognition, however, can be a cumbersome task. Hidden Markov models (HMMs), a pattern recognition technique commonly used in speech recognition, can be used for recognizing certain classes of gestures. Existing HMM toolkits for speech recognition can be adapted to perform gesture recognition, but doing so requires significant knowledge of the speech recognition literature and its relation to gesture recognition. This paper introduces the Georgia Tech Gesture Toolkit GT2k which leverages Cambridge University's speech recognition toolkit, HTK, to provide tools that support gesture recognition research. GT2k provides capabilities for training models and allows for both real--time and off-line recognition. This paper presents four ongoing projects that utilize the toolkit in a variety of domains.