Lecture 9 | MIT 6.832 (Underactuated Robotics), Spring 2018

This lecture provides a solid introduction to trajectory optimization, a fundamental tool in robotics and control. The content is well-structured, building on previous lectures and clearly explaining the motivation for moving from global methods to local trajectory optimization. The instructor’s expertise is evident, and the presentation is clear, with mathematical formulations and practical examples. The main strength is the pedagogical approach: Tedrake contrasts trajectory optimization with dynamic programming and sum-of-squares methods, highlighting the trade-offs in scalability and optimality. He also discusses the practical aspects, such as the fragility of non-convex solvers and the importance of initialization. The lecture is technically accurate and aligns with standard practices in the field. However, it lacks explicit citations to external sources, which is typical for a lecture but limits the ability to verify specific claims. The use of a live demonstration, including a failure case, adds authenticity and illustrates the challenges in practice. The content is suitable for an advanced undergraduate or graduate audience with a background in control theory and optimization. Overall, this is a high-quality educational resource that effectively conveys the key concepts and practical considerations of trajectory optimization.

The title accurately reflects the content: a lecture on underactuated robotics, specifically focusing on trajectory optimization.

8/10

Lecture by MIT professor Russ Tedrake, part of a well-established course. Content is technically rigorous, based on established optimization and control theory. No citations provided in the video, but the course website offers additional resources. The lecture is a primary educational source, not peer-reviewed, but highly reliable for its domain.

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

• 00:00Introduction and recap of previous lectures on dynamic programming and sum-of-squares methods.
• 02:00Discussion of the limitations of dynamic programming and sum-of-squares for high-dimensional systems.
• 05:00Introduction to trajectory optimization as a local method for a single initial condition.
• 10:00Formulation of trajectory optimization for linear discrete-time systems as a quadratic program.
• 15:00Discussion of alternative cost functions, such as L1 norms, leading to linear programming.
• 20:00Comparison with LQR and the extension to constrained problems.
• 25:00Demonstration of trajectory optimization on a pendulum swing-up and cart-pole system.
• 30:00Discussion of the fragility of non-convex solvers and the importance of initialization.
• 35:00Conclusion and pointers to course website for further materials.

• Underactuated Robotics Course Website — Referenced in the video description as the course website for additional materials.

• Underactuated Robotics Course Website — The course website provides lecture notes and additional resources that align with the content of this lecture.