How to Control What You Can't See [video]
This video explores strategies for managing factors beyond one's direct perception or control, such as hidden variables in systems, uncertainty in decision-making, and unseen influences. It discusses how to identify blind spots, use indirect measurement, and adopt a proactive mindset to navigate unpredictable environments effectively.
Background
- The video explores the "control barrier function" framework, a mathematical method used in control theory to keep autonomous systems (drones, robots, self-driving cars) operating within safe limits — even when some variables can't be directly measured.
- This builds on decades of research in control theory and Lyapunov stability, but CBFs (formalized around 2014-2017 by Aaron Ames and others) give engineers a practical way to guarantee safety without sacrificing performance.
- The key insight: you can enforce constraints (like "don't crash" or "stay in lane") by translating them into a mathematical condition that overrides the primary controller only when necessary, similar to how a human driver might brake sharply to avoid an obstacle while still steering normally the rest of the time.
- Why it matters: as robots and autonomous systems leave carefully controlled labs and enter messy real-world environments, they need formal guarantees — not just statistical likelihoods — that they won't violate safety boundaries. CBFs are being adopted in self-driving cars, quadrotor drones, and even assistive medical devices.