David Feil-Seifer: Decision-making architecture in robotics

Title

Simultaneous Task Collaboration Architecture

Mentor

David Feil-Seifer

Department

Computer Science & Engineering

Biosketch

David Feil-Seifer is an Associate Professor in Computer Science at the University of Nevada, Reno. His primary research interests are Socially Assistive Robotics (SAR) and User Interface design for Unmanned Autonomous Systems (UAS-UI). His research is motivated by the potential for SAR to address health-care crises that stem from a lack of qualified care professionals for an ever-growing population in need of personalized care as well as the uses for aerial robots for disaster mitigation.

Prior to his tenure at UNR, he was a Postdoctoral Associate at Yale University at the Social Robotics Lab under the direction of Prof. Brian Scassellati. He was awarded a National Science Foundation/Computing Research Association Computing Innovation Fellowship to support his postdoctoral work. He received a B.S. in Computer Science from the University of Rochester and a M.S. and Ph.D. in Computer Science from the Viterbi School of Engineering (VSoE) at the University of Southern California (USC) under the direction of Prof. Maja Matarić. He helped coin the term, Socially Assistive Robotics (SAR), and studied its applications for children with autism spectrum disorders (ASD). He has been awarded the Mellon award for Mentoring, the Order of Arete, and the USC VSoE Best Dissertation award.

His research is motivated by the potential for socially assistive robotics (SAR) to address health-care crises that stem from a lack of qualified care professionals for an ever-growing population in need of personalized care. SAR provides assistance through social rather than physical interaction and can augment human caregivers. He has worked to develop SAR algorithms and complete systems that are relevant to domains such as post-stroke rehabilitation, elder care, and therapeutic interaction for children with autism spectrum disorders (ASD). The key challenge for such autonomous SAR systems is the ability to sense, interpret, and properly respond to human social behavior, especially given the unpredictable and heterogeneous nature of human responses. He has published 14 papers with undergraduates in conferences and journals.

Project Overview

There are a variety of implementations possible for decision-making architectures in robotics. These implementations allow for a robot, or multiple robots, to decide the most efficient way to complete a task. One type of implementation is hierarchical task networks, which are networks composed of primitive nodes, goal nodes, and compound nodes. The architecture that is the basis of this work organizes a task and its constraints into hierarchical task trees, which are composed of the same nodes as a hierarchical task network.

When completing a task, a robot must adhere to these constraints. If the robot were to make a sandwich, it would first place the bread, then place the meat or cheese, then place a second slice of bread. Currently, the architecture understands the constraints of THEN, AND, and OR:

THEN: Sequential Ordering - first one child, then the other
AND: Nonsequential Ordering - complete the children in any order
OR: Alternate ordering - complete any of the children
The focus of this research is to implement a new constraint, WHILE. WHILE will enforce collaborative behavior; while holding one child, complete another child.