Overview
This course focuses on the challenges of designing robots to work alongside humans in real-world scenarios. The learning outcomes include understanding how robots can learn from natural human interactions and predict human intent. The course teaches skills such as tackling feedback-driven covariate shift and planning with learned forecasts using a graph neural network approach. The teaching method involves a seminar-style talk diving into two core challenges in interactive imitation learning. The intended audience for this course includes researchers, practitioners, and students interested in machine learning, robotics, and artificial intelligence.
Syllabus
Intro
Welcome
The Question
Two Fundamental Challenges
Aurora Driver
Programming Rules
Markov Decision Process
Challenges
Feedback Drives Covariate Shift
How common is this problem
Feedback driving covariate shift
Benchmarks
Infinite Data Limit
Hard Setting
Dagger
Interactive Expert
Expert Intervention Learning
Quantitative Plots
NonRealizable Expert
Simulation
Question Querying
Driving Simulators
Open Questions
Example
Grammar Modes
Merging Scenario
Transformer Net
Conclusion
Taught by
Stanford Online