Explore crowd dynamics modeling through reinforcement learning in this 35-minute Strange Loop Conference talk. Delve into the challenges of traditional approaches and discover how reinforcement learning offers a more robust solution for diverse environments. Learn about extended Kalman filters, implementing PPO in Unity-ML, and result validation. Examine the interplay between unexpected and predictable crowd behaviors, providing insights into this rich domain of modeling and simulation. Follow speaker Tomas Diaz's journey from Wall Street to innovative software solutions, focusing on graphics, machine learning, and virtual reality applications in extreme environments.
Overview
Syllabus
Intro
THE PROBLEM
PEOPLE ARE SUMMATIONS OF SOCIALITY FORCES - 1990
REINFORCEMENT LEARNING
THE DIFFICULTIES
WHAT IS PPOT
UNITY ML-AGENTS
ONE CODE EXAMPLE
KALMAN'S GOAL
KALMAN FILTER DEMO
WHY DID I TELL YOU THAT
FINAL DEMO
CLOSING THOUGHTS
Taught by
Strange Loop Conference