In this work we present a conceptual design to automatically evaluate a subject's performance for a homebased stroke rehabilitation system. We propose to model a reaching task as a trajectory in the state space of hand part features and then use reward learning to automatically generate new ratings for subjects to track performance over time. Neuromuscular rehabilitation of the upper-extremity after a stroke requires dedicated hours of arm and hand exercises with a therapist. Often to improve the flexibility of the hands, therapists ask patients to manipulate differently shaped objects and move them around. We are working towards developing a home-based neuromuscular rehabilitation system Figure 1: An ideal home based therapy system using a single by doing away with markers and using 2D computer vision camera where a subject can do upper-extremity exercises by instead. As a first step, we are able to identify different grasp manipulating objects in the home on a simple kitchen table.

We are looking at new ways of building algorithms for synthesizing and rendering animation in social robots that can keep them as interactive as necessary, while still following on principles and practices used by professional animators. We will be studying the animation process side by side with professional animators in order to understand how these algorithms and tools can be used by animators to achieve animation capable of correctly adapting to the environment and the artificial intelligence that controls the robot. Figure 1: Two example scenarios featuring a touch-based Robotic characters are becoming widespread as useful multimedia application, sensors, and different robots.

The EMPOWER project creates opportunities for physically disabled individuals (namely quadriplegics) to operate robots through the web browser. Robotic technology has the ability to unlock productivity and grant greater purpose to mentally capable, but physically disabled, users. The goal of the EMPOWER project is to foster independence for the physically disabled by lowering barriers such as accessibility and cost. The potential of the EMPOWER project can best be illustrated through the prototyping projects between Mr. Evans and Brown University. Through our web-enabled AR.Drone (running ROS), Mr. Evans has been able to engage in activities in Providence, RI from his home in California, where he remotely pilots AR.Drones. The live video feed from the embedded cameras in the quadricopter provide Mr. Evans a vehicle to explore and interact with people and places far beyond the confines of his bed.

CoBot is a service mobile robot that has been continuously Frame: GoTo deployed for extended periods of time in a multi-floor - Parameters: destination office-style building (Biswas and Veloso 2013). While moving in the building CoBot, is able to perform multiple tasks Frame: DeliverObject for its users; the robot is able to autonomously navigate to - Parameters: object, source, destination any of the rooms in the building, to deliver objects and messages and to escort visitors to their destination. While apparently Figure 1: Semantic frames of the two task CoBot is able to very different, all the tasks CoBot is able to perform execute from spoken commands. Often, only being able to move, is not enough to accomplish the task required; if CoBot needs to deliver an object, given that it does not have arms, it cannot pick it up by itself, similarly when it needs to travel across floors it cannot push the elevator button. In order to overcome its limitation CoBot ask for help to humans, either the user or bypasser, achieving symbiotic autonomy (Rosenthal, Biswas, and Veloso 2010).

Research on POMDP formulations for collaborative tasks in game AI applications (Nguyen et al. 2011; Macindoe, The development of new industrial robotic systems that operate Kaelbling, and Lozano-Pérez 2012; Silver and Veness in the same physical space as people highlights the 2010) also assumed a known human model. Additionally, emerging need for robots that can integrate seamlessly into previous partially observable formalisms (Ong et al. 2010; human group dynamics by adapting to the personalized style Bandyopadhyay et al. 2013; Broz, Nourbakhsh, and Simmons of human teammates. This adaptation requires learning a statistical 2011; Fern and Tadepalli 2010; Nguyen et al. 2011; model of human behavior and integrating this model Macindoe, Kaelbling, and Lozano-Pérez 2012) in assistive into the decision-making algorithm of the robot in a principled or collaborative tasks represented the preference or intention way. We present a framework for automatically learning of the human for their own actions, rather than those of human user models from joint-action demonstrations the robot, as the partially observable variable.

Vocalics is the study of the nonverbal aspects of speech, such as volume, pitch, and rate. Our contribution is a parametric We present an overview of the control, recognition, decision-making, vocalic behavior controller that autonomously adjusts and learning techniques utilized by the Interaction the robot speaker volume based on models of how a Lab (robotics.usc.edu/interaction) at the University human user will hear speech produced by the robot. These of Southern California (USC) to enable autonomy in sociable models vary with distance, orientation, and perceived environmental and socially assistive robots. These techniques are implemented interference (Mead & Matarić 2014). Our future with two software libraries: 1) the Social Behavior work will investigate adapting the pitch and rate of speech Library (SBL) provides autonomous social behavior produced by a robot to improve user speech perception.

In this work we present a method that allows to learn a cost function for motion planning of human-robot collaborative manipulation tasks where the human and the robot manipulate objects simultaneously in close proximity. Our approach is based on inverse optimal control which enables, considering a set of demonstrations, to find a cost function balancing different features. The cost function that is recovered from the human demonstrations is composed of elementary features, which are designed to encode notions such as safely, legibility and efficiency of the manipulation motions. We demonstrate the approach on data gathered from motion capture of human-human manipulation in close proximity of blocks on a table. To demonstrate the feasibility and efficacy of our approach we provide initial test results consisting of learning a cost function and then planning for the human kinematic model used in the learning phase.

A study of long-term interaction with the robot embodiment of the companion called Sarah was conducted during the summer of 2012. The aim of the study was to see long-term implications when the robot embodiment was in a natural setting. The robot interacted with 5 participants for 3 weeks in a office environment running continuously.

Over the last five years, and while developing an architecture for autonomous service robots in human environments, we have identified several key decisional issues that are to be tackled for a cognitive robot to share space and tasks with a human. We introduce some of them here: situation assessment and mutual modelling, management and exploitation of each agent (human and robot) knowledge in separate cognitive models, natural multi-modal communication, "human-aware" task planning, and human and robot interleaved plan achievement. As a general "take home" message, it appears that explicit knowledge management, both symbolic and geometric, proves to be a successful key while attempting to address these challenges, as it pushes for a different, more semantic way to address the decision-making issue in human-robot interactions.

Touch can be a powerful means of communication especially when it is combined with other sensing modalities, such as speech. The challenge on a humanoid robot is to sense touch in a way that can be sensitive to subtle cues, such as the hand used and amount of force applied. We propose a novel combination of sensing modalities to extract touch information. We extract hand information using the Leap Motion active sensor, then determine force information from force sensitive resistors. We combine these sensing modalities at the feature level, then train a support vector machine to recognize specific touch gestures. We demonstrate a high level of accuracy recognizing four different touch gestures from the firefighting domain.