The Future of Machine Learning in Robotics: Meghan Dailey, Alumni Award Winner
Meghan Dailey, a machine learning specialist on the Scientific Computing and Research Consulting team at the University of Michigan’s Advanced Research Computing division, discusses machine learning in robotics and research collaborations at the University of Michigan.
In this talk
- 1:57
Machine learning depends on fit-for-purpose data
Dailey describes robotics as a combination of controls and dynamics, perception, learning, and intelligence. For machine learning to help, its data needs to match the question being asked and cover the situations a robot may encounter.
- 3:44
A path from underwater robotics to social robotics
Dailey traces her interest in robotics from computer engineering at Santa Clara University, where she worked on mapping Lake Tahoe’s floor and collecting data from abandoned ocean boreholes. At Michigan, she studied whether biometric signals could help social robots respond to people’s emotional cues, including interest during a presentation.
- 13:14
Research computing projects put models to work
At Michigan’s Advanced Research Computing division, Dailey advises researchers on machine learning and works with teams using the Great Lakes cluster. One project uses a ResNet model to detect and track fish in sonar video from wetlands near Midland, Michigan; another identifies kidney structures in medical scans to support studies of kidney health.
- 20:40
Applications shape machine learning’s role in robotics
Dailey argues that useful progress can come from applying existing machine learning methods to new questions, across areas such as computer vision, language processing, and reinforcement learning. She sees robots and people working together as a key direction, with machine learning helping robots respond in ways people can work with.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.