Description:

Title:
Designing a Robot's Interaction with the Environment

Abstract:
A robot's behavior is the resulting interaction between that robot and the environment. However real world environments, like our homes and offices, are cluttered, complex, and changing dynamically. Here, the environment can not be directly controlled and the full state of the environment will never be know. Therefore the robot's behavior may only be controlled indirectly through the design of the robot's systems: the hardware platform, software controllers, and sensor feedback. This talk will present new models of the mechanics of a robot's interaction with the environment and study structural properties of those models gain design insight into the robot's systems. In particular, divergence metrics over the robot's dynamics are identified to measure how uncertainty in the system will evolve over time. These metrics are then used in planning the robot's controllers so that the resulting interaction with the environment is more robust to uncertainty. These ideas are applied to both a home-assistance manipulation robot and an outdoor rough terrain legged robot.

Bio:
Aaron M. Johnson is currently a Postdoctoral Fellow in the Robotics Institute at Carnegie Mellon University. His research interests include novel robot design - behavior design, controller design, and platform design - as well as dynamic transitions, physics-based planning, bio-inspired robotics, actuator modeling, and robot ethics. Dr. Johnson received his Ph.D. degree from the Department of Electrical and Systems Engineering at the University of Pennsylvania in 2014, and his B.S. degree from the Department of Electrical and Computer Engineering at Carnegie Mellon University in 2008. He was formerly a Visiting Researcher with Boston Dynamics, Electrical Engineering Intern at iRobot, and a Research Assistant with the Biorobotics Lab at Carnegie Mellon University. He was a Best Student Paper Finalist at the IEEE International Conference on Robotics and Automation in 2013, as well as at the Climbing and Walking Robots Conference in 2012.