Early experiences with engineering sprang from a natural knack for problem-solving and the invigorating thrill of riding roller coasters. I was intrigued by human-machine interaction, curious how technology could create vivid sensations of flight, danger, joy, and adventure. Robots, in particular, piqued my interest. Their ability to mimic human behaviors afforded robots the unique opportunity to offer deeper, more meaningful connections: collaboration, care, and teaching. Having interacted with a variety of robots, particularly audio-animatronics at theme parks, I noticed a striking limitation to this potential: my interaction with the robot was always remote. Despite their realistic movement and expression, robots were still far from accurately understanding and responding to the complexities of

human moition, our delicacy, and response given uncertainty. Harnessing safe, physical interaction between robots and humans would provide a fascinating new level of interaction, with potential to impact the fields of medicine, defense, education, and manufacturing. I took on the challenge of bringing humans and robots together, experimenting with robot kits early, pursuing math and science however I could. Left hungry by the absence of engineering instruction available to students through middle and high school, I vowed to promote engineering outreach, creating the difference I wanted to see.


With Disney's WALL-E Audio-Animatronic

Background Experience
Exposure to research and industry offered challenging opportunities for growth and learning, ultimately shaping my decision to pursue graduate studies in Mechanical Engineering. Early internships within The Boeing Company exposed me to the business applications of engineering as I worked with diverse teams to identify problems and implement solutions; I developed important skills while interacting with a variety of customers. At school, I found engaging technical challenges through independent study and graduate-level experimental robotics courses. I explored my interest in robotics through simulation research at Walt Disney Imagineering Research and Development. As I currently define my research direction, I am fortunate to be able to draw upon this diverse set of skills and experiences.

Motivation
Robots have great potential for assisting impaired individuals, from enhancing their natural abilities to enabling independent activities of daily living. However, safety considerations limit the feasibility of physical interaction between humans and robots. The field of human-robot interaction (HRI) lacks a mechanical foundation on which to design robots that can accurately understand and safely respond to these physical interactions. Research in the field of HRI has thus largely avoided touch in favor of psychosocial, linguistic, and cognitive research (e.g. socially-assistive robotics), with little use of physical interaction. A promising foundation for physical HRI is the design of

robots that guide human movement, as it provides a tangible context for physical interaction and incorporates many critical modes of touch: force sensing, application, and reaction. Such robots would help users navigate obstacles while maintaining proper balance. A physically guiding robot offers two key technical challenges:

1. accurate sensing of natural human dynamics, and
2. responsive application of forces that successfully guide human movement to a desired target.

As a basis for physical HRI, guiding would offer insights into human mechanics (e.g. balancing) and expand the scope of current research in the field, leading the development of other safe, touch-friendly robots.


Guided by Honda's ASIMO Robot at Stanford

Hypothesis
A mechanical understanding of natural human guiding can be modeled and applied to the development of a robot control and sensor scheme that accurately detects changes in impedance and intelligently guides human motion by safely applying required forces. The following figure shows a block diagram of the proposed dynamic response using computational models.


Block diagram of guiding robot dynamic response

Proposed Method
Research will be conducted in two phases. First, I plan to study natural physical human interactions to describe how humans dynamically apply and react to guiding forces. I plan to develop a computational model of safe guiding mechanics that defines how guiding parties adjust their impedance to achieve task objectives. In the model, I will consider current neuroscience research to properly understand human neuromotor function (e.g. force escalation). To define the model, I will conduct physical human studies using various force-sensing and optical tracking techniques to quantify subject motion and kinetics. Preliminary tests will examine simple guiding interactions, with human users guiding each other through virtual, planar paths with variable instability. By altering the guides’ knowledge of instability along the desired path, I intend to measure how information alters their physical response while they are guiding. Subsequent tests will use instrumented gloves and gradually add complexity, as I plan to examine more intricate paths, different guiding points on the body, and more scenario-driven guiding (e.g. avoiding obstacles, sitting, walking).

The second phase is the implementation of a controller for impedance-matching robots, described in the figure to the left. Sensed human forces and position would be used with

computational impedance models developed in the previous phase. The mathematical model relating instability, force, and position, will estimate the user’s impedance and inform whether that of the robot should match or slightly differ. The robot impedance and relative position error yield the force it must apply to the user, and the stiffness with which it should do so. Initial controller tests will use a six-degree-of-freedom robotic arm (e.g. PUMA) with which users will remotely interact via haptic devices. Once the controller is validated remotely, force sensors will be affixed to the robot’s guiding hand for studies involving direct contact with subjects, testing simple pushing exercises first. These tests will mimic those from the previous phase, replacing human guides with a robot. Final controller iterations will be tested on a humanoid robot with more sophisticated instrumentation, such as ASIMO. The final control need not mimic human guiding, but rather, produce forces sensitive to human reaction.

Intellectual Merit
The proposed control scheme for guiding human motion will contribute to our understanding of human dynamic response. More than just develop a novel way of guiding, this research can advance the field of robotics by providing foundational insights into other forms of physical human-robot interaction. At a

basic level, this research addresses the field’s need for smart, safe, touch-friendly robots. With the understanding of this potentially transformative work, other scientists in the field could apply the guiding control scheme to existing research such as surgical robotics, or branch into new fields of physical HRI, such as dance, rehabilitation, and childcare. In addition, Stanford offers unique technological resources critical for this research. This work will be conducted under Dr. Allison Okamura, who offers expertise inhuman-machine systems capable of haptic interaction and a variety of teleoperation devices to be used for preliminary controller testing (e.g. Sensable’s Phantom Premium Robot). Her lab is certified by the Institutional Review Board for human subject research, and I have completed the associated training. I will also benefit from access to humanoid robots in other departments, such as ASIMO and locally (e.g. Willow Garage’s PR2) for final controller implementations and testing.

Broader Impact
Results of this work could foster independence for impaired individuals. For instance, guiding robots could aid patients with visual impairment. According to 2010 and 2004 studies by the World Health Organization and U.S. National Institutes of Health, respectively, an estimated 285 million people (4% of the world’s population) are visually impaired, including over

3.3 million Americans, and with an expected increase of 60% by 2020. In addition to assisting the visually impaired, guiding robots could be used to safely assist people with other mental, physical, and sensory impairments. Article 26 of the 2006 United Nations Convention on the Rights of Persons with Disabilities affirms the need to “take effective and appropriate measures [to enable] full physical…ability” for the world’s estimated 650 million disabled people. Responsive robots capable of physical HRI could not only enhance physical ability when the robot is in use, they could also train individuals to maintain balance, navigate around obstacles, and complete crucial, day-to-day tasks. Using models describing human guiding dynamics, a guide robot could possess the intelligence needed to identify the necessary assistance and utilize the required sensitivity. Additional applications include assistance with skilled tasks and robotic surgery. Opportunities for outreach and education are distinct and engaging: users learn about robotics and engineering as they touch, interact and are guided.

Work is underway defining and implementing my first experiment. I plan to update this page with information about the experiments, including results, videos, and papers describing my work. This page will also contain links to my publications, so stay tuned with forthcoming updates!