Lecture 16: MIT 6.832 Underactuated Robotics (Spring 2022) | "Hybrid Trajectory Optimization"

This lecture is a masterclass in hybrid trajectory optimization, delivered with exceptional clarity and depth. The instructor, Russ Tedrake, is a leading authority in the field, and his expertise shines through in the careful progression of ideas and the emphasis on practical computational methods.

The lecture builds on previous material, establishing the necessary theoretical foundation before diving into numerical techniques. The review of limit cycle stability and hybrid dynamics is concise but sufficient, ensuring that students have the context needed to appreciate the optimization formulations that follow.

One of the key strengths is the explicit treatment of time as a decision variable in trajectory optimization. This is a subtle but crucial point, as it allows the optimizer to stretch or shrink the trajectory duration to satisfy periodicity constraints. The instructor’s explanation of why this is necessary, using the Van der Pol oscillator as an example, is particularly illuminating.

The demonstration of finding limit cycles via direct collocation is both practical and instructive. The instructor shows actual code snippets and discusses the importance of initialization, noting that a trivial solution (the fixed point) exists and must be avoided by providing a good initial guess. This practical insight is invaluable for anyone attempting to implement these methods.

The transition to the rimless wheel, a canonical hybrid system, is handled seamlessly. The instructor clearly explains how to incorporate the reset map and periodicity constraints into the optimization, effectively turning the problem into a feasibility problem. This example serves as a template for tackling more complex walking robots.

The lecture also includes a thoughtful response to a student question about region of attraction estimation, demonstrating the instructor’s ability to connect the material to broader challenges in nonlinear control.

Overall, this lecture is rigorous, well-structured, and highly informative. It strikes an excellent balance between theory and practice, making it suitable for graduate students and researchers in robotics and control. The only minor criticism is that the lecture assumes prior knowledge of trajectory optimization and hybrid systems, so it may not be accessible to complete beginners. However, for its intended audience, it is outstanding.

The title accurately reflects the content: a lecture on hybrid trajectory optimization for underactuated robotics, covering limit cycle finding and optimization for systems with contact.

9/10

Lecture from MIT's Underactuated Robotics course, presented by a leading expert in the field. The content is rigorous, well-structured, and based on established methods in trajectory optimization and hybrid systems. The lecture is part of a reputable academic series, and the technical depth is high, with clear explanations and mathematical formulations.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and recap of limit cycle stability and hybrid dynamics.
• 05:00Discussion on finding limit cycles via simulation vs. optimization.
• 10:00Formulation of trajectory optimization for limit cycles, emphasizing time as a decision variable.
• 15:00Direct collocation example for Van der Pol oscillator, showing code and results.
• 20:00Student question about region of attraction and response.
• 25:00Transition to rimless wheel: incorporating reset map and periodicity constraints.
• 30:00Detailed formulation of rimless wheel optimization, including constraints.
• 35:00Discussion on generality of the approach for other systems.

• Underactuated Robotics (course textbook) — The lecture is based on the course textbook, which provides comprehensive material on trajectory optimization and hybrid systems.

• Underactuated Robotics (course textbook) — The lecture content aligns with the course textbook, which covers similar topics in depth.