Mini-Lecture 9 (Trajectory Optimization) | MIT 6.832 (Underactuated Robotics), Spring 2021

This lecture provides a solid introduction to trajectory optimization, a core topic in robotics and control. The instructor, Russ Tedrake, is a leading expert in the field, and the content is well-structured, building from the fundamental motivation to specific numerical methods. The key strength is the clear explanation of why trajectory optimization is a practical approach: it avoids the curse of dimensionality by focusing on a single trajectory, making it scalable to high-dimensional systems. The discussion of various methods (direct transcription, shooting, collocation, pseudo-spectral) gives a comprehensive overview, and the instructor’s comments on their relative merits are insightful. The use of examples, such as autonomous driving, helps to ground the concepts in real-world applications. The lecture also touches on important practical considerations, such as the use of slack variables and the choice of solvers, which are often glossed over in introductory treatments. However, the lecture is not without limitations. As a mini-lecture, it covers a lot of ground in a short time, and some topics, such as pseudo-spectral methods, are only briefly mentioned. The discussion of solvers is somewhat dated, as it focuses on SNOPT and does not mention more recent developments like IPOPT or the use of CasADi. Additionally, the lecture assumes a certain level of familiarity with optimization and control theory, which may make it less accessible to beginners. The Q&A segment is useful but could have been more extensive. Overall, this is a high-quality educational resource that effectively conveys the key ideas of trajectory optimization, though it is not a substitute for a comprehensive textbook or research paper.

The title accurately reflects the content: a mini-lecture on trajectory optimization within the MIT course on underactuated robotics.

8/10

Lecture from MIT OpenCourseWare by a renowned expert in robotics, based on established optimization methods and supported by course notes. The content is technically accurate and well-structured, though it is a lecture rather than peer-reviewed research.

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

• 00:00Introduction and recap of the lecture on trajectory optimization.
• 01:00Discussion of the main idea: focusing on a single trajectory to avoid exponential state-space complexity.
• 03:00Explanation of the general formulation of trajectory optimization with running costs.
• 05:00Overview of direct transcription, shooting, collocation, and pseudo-spectral methods.
• 07:00Q&A on slack variables and their role in optimization.
• 10:00Discussion of solvers like SNOPT and the trend toward augmented Lagrangian methods.
• 12:00Application of trajectory optimization in autonomous driving, including lane changes and safety constraints.
• 15:00Comparison with sum-of-squares and the importance of online planning.