Mini-Lecture 10 (Trajectory Stabilization) | MIT 6.832 (Underactuated Robotics), Spring 2021

This lecture provides a rigorous and insightful treatment of trajectory stabilization for underactuated systems. The instructor, presumably Russ Tedrake, demonstrates deep expertise and pedagogical clarity. The content is well-structured, building from the fundamental problem of open-loop instability to the elegant solution of time-varying LQR, and then extending to Lyapunov-based certification. The use of the perching example throughout effectively ties together the concepts, illustrating how trajectory optimization, LQR, and sums-of-squares verification can be integrated to produce a robust controller with formal guarantees. The mathematical derivations are sound, and the instructor takes care to explain the intuition behind the equations, such as the time-varying Riccati equation and the interpretation of the cost-to-go as a Lyapunov function. The discussion of alternative methods like MPC provides a broader context and highlights trade-offs. The lecture is highly technical and assumes prior knowledge of control theory and optimization, but it is delivered in an accessible manner for an advanced audience. The sources are primarily the course notes and the instructor’s expertise, which are credible in this context. The title accurately reflects the content. Overall, this is an excellent educational resource for students and practitioners in robotics and control.

The title accurately reflects the content: a mini-lecture on trajectory stabilization within the context of underactuated robotics.

8/10

Lecture from MIT OpenCourseWare, presented by an expert in the field, with rigorous mathematical derivations and references to course notes. The content is well-structured and based on established control theory.

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

• 00:00Introduction and recap of the lecture's core ideas: trajectory stabilization using LQR, linearization along a trajectory, time-varying Riccati equation.
• 02:00Discussion of why open-loop playback is unstable and the need for feedback control.
• 05:00Introduction of time-varying LQR: linearization in a moving coordinate system, time-varying A(t) and B(t), and solution via backward Riccati equation.
• 10:00Alternative stabilization methods: model predictive control (MPC) and linear MPC for constrained linear systems.
• 15:00Using the cost-to-go from LQR as a Lyapunov function for certification, and the concept of invariant sets and funnels.
• 20:00Detailed explanation of the perching example: trajectory optimization, LQR, and sums-of-squares Lyapunov analysis to compute a funnel of initial conditions.
• 25:00Student question about the relationship between LQR and the Lyapunov function; instructor explains the time-invariant case first.
• 30:00Extension to time-varying Lyapunov functions and the condition for staying inside the funnel (v_dot <= rho_dot).
• 35:00Discussion of what happens outside the certified region and the notion of inner approximation.
• 40:00Conclusion and wrap-up, emphasizing the integration of methods.

• MIT 6.832 Underactuated Robotics Course Notes — The instructor references the course notes, which contain expanded material on these topics.

• Underactuated Robotics Course Notes — The lecture is based on the course notes, which provide detailed derivations and additional examples.