Better Ways to Measure and Move - Joint Optimization of Agent's Physical Design and Computational Reasoning

Explore cutting-edge research on joint optimization of physical design and computational reasoning for intelligent agents in this hour-long lecture by Matthew Walter from Toyota Technological Institute at Chicago. Delve into innovative learning-based methods for automated, data-driven optimization of sensor networks and physical configurations alongside computational inference and control. Discover frameworks for optimizing sensor network design, legged robot structures, and soft robot designs coupled with control policies. Gain insights into improving test-time generalization of learned policies and the potential impact on various robotic applications, from underwater vehicles to self-driving cars. Examine the speaker's background in intelligent, perceptually aware robots and his contributions to the field of machine learning-based solutions for robust robotic interactions in unstructured environments.

Introduction
Welcome
What is TTIC
TTIC Hiring
Perception Lab
Sensor Design
Beacon Based Localization
Beacon Placement
Design and Reasoning
Evolutionary Methods
Design Control
OpenAI Gym
Design Evolution
Results
Design Distribution
Soft Robotics
Design Space
Pneumatic Localization
Design Spaces
Modeling Dynamics
Problems
Recent Work
Domain Randomization
Intuition
AI Driving Olympics
Acknowledgements
Python Skit
Discussion