PhD Candidate
University of Colorado Boulder
Areas of Interest
- Artificial Intelligence
- Control, Intelligent Systems, and Robotics
Poster
Control and Learning on Edge Devices for Peripheral Robot Intelligence
Abstract
Advances in microelectronics and machine learning allow us to develop devices capable of learning over their lifetime without requiring constant communication to a cloud just like human guts have an entire nervous system independent of the brain and the spinal cord called the enteric nervous system (ENS). I’m interested in developing on-board control and estimation policies capable of decoupling the computation required to make quick decisions at the location of the subsystem instead of asking a central computer what it needs to do. This requires lightweight and cheap estimation and control methods that are highly effective. To this end we develop multiple open source tools that allow for cheap in-situ estimation and fast online nonlinear controls that are informed by an embedded neural network. The next steps of this work is to develop control policies capable of adapting to novel environments without having been specifically trained for novelty. This process of life-long learning may enable edge technologies such as wearable devices, advanced prostheses, mobile phones and other interfaces to adapt to a much larger range of use cases in our open-ended, unstructured world.
Bio
Sarah is a fourth year PhD student in computer science working at the Correll Lab of Robotic Materials under Nikolaus Correll at the University of Colorado Boulder. Her primary interest is creating tools to facilitate peripheral intelligence in robotic systems through microcontrollers. She previously completed her BS and MS in 2015 and 2017 respectively in aerospace engineering at Penn State University.
