6.8210 Spring 2024 Lecture 10: Trajectory Optimization I

The lecture provides a rigorous and well-motivated introduction to trajectory optimization, a cornerstone of modern robotics and control. The instructor, Russ Tedrake, is a leading expert in the field, and his presentation reflects deep understanding and pedagogical clarity. The content is scientifically sound, building on established mathematical foundations. The lecture begins by situating trajectory optimization within the broader context of optimal control, contrasting it with dynamic programming and Lyapunov methods. This framing is valuable as it clarifies the trade-offs between different approaches. The core contribution is the formulation of the linear discrete-time trajectory optimization problem as a quadratic program. This is presented with precision, clearly defining decision variables, constraints, and the objective function. The instructor correctly emphasizes that the ability to incorporate constraints is the key advantage over LQR, which is limited to unconstrained problems. The lecture also touches on important practical aspects, such as the reliability of convex solvers and the potential for infeasibility. The presentation is well-structured, with a logical flow from motivation to formulation to discussion of implications. The use of a concrete example (though not fully developed in this excerpt) helps to illustrate the concepts. The lecture is aimed at graduate students with a background in control and optimization, and it assumes familiarity with linear algebra and dynamic programming. The technical level is appropriate for the target audience. The lecture does not include any apparent biases or unsupported claims. The sources cited are likely standard textbooks and papers in the field, though specific references are not mentioned in the transcript. Overall, this is an excellent lecture that provides a solid foundation for further study in trajectory optimization. The only minor criticism is that the lecture is somewhat introductory and does not delve into the numerical methods for solving QPs or the extensions to nonlinear systems, but these are likely covered in subsequent lectures.

The title accurately reflects the content: the lecture introduces trajectory optimization, focusing on the linear discrete-time case and its formulation as a quadratic program.

9/10

Lecture from MIT OpenCourseWare, presented by a recognized expert in robotics and control (Russ Tedrake). The content is rigorous, well-structured, and based on established mathematical methods. The lecture is part of a formal course, ensuring high reliability.

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

• 00:00Introduction and course logistics, midterm announcement.
• 02:00Review of dynamic programming and its limitations, curse of dimensionality.
• 05:00Motivation for trajectory optimization: focusing on a single initial condition.
• 08:00Formulation of linear discrete-time trajectory optimization problem.
• 10:00Introduction of quadratic program (QP) formulation and its advantages.
• 12:00Comparison with LQR and discussion of constraints.
• 14:00Discussion of convex solvers and infeasibility.