Course on Pontryagin's Maximum Principle

Instructor

Andrew Lewis
Department of Mathematics and Statistics
Queen's University
andrew@mast.queensu.ca

Lectures: May 9-16 2006.



Downloads

Some version of these notes were distributed to students at the beginning of the course. However, they have been significantly revised since the original distribution. If you have a version of these notes with a date earlier than 16/05/2006, please download a later version. Earlier versions had aggredious errors. I will continue to make modifications to these until I stop doing so at some undetermined time.

[Course notes]

Course description

This is a short course (maybe 20 hours) offered to graduate students at Universitat Politècnica de Catalunya.

The intent of the course is to introduce students to Pontryagin's Maximum Principle. The structure of the course is as follows.

  1. We begin by covering a few topics from the calculus of variations: the Euler-Lagrange equations, the Legendre condition, and the Weierstrass excess function.
  2. An attempt will be made to draw a line from the calculus of variations to the Maximum Principle through the so-called Skinner-Rusk formulation of the calculus of variations.
  3. We clearly define control systems, classes of trajectories, and the various optimal control problems we consider. We then clearly state the Maximum Principle.
  4. To prove the Maximum Principle, we take a big detour through topics such as control variations and the relationship between these and the reachable set. By doing this we will indicate how the Maximum Principle is not just useful in optimal control, but in control theory in general.
  5. We prove the Maximum Principle.
  6. We discuss some aspects of the Maximum Principle that are really only meaningful after one understands the proof.
  7. The Maximum Principle will be used to say some things about linear quadratic optimal control and linear time-optimal control.

Prerequisite: Basic real analysis and differential equations are the only things that will really be essential prerequisites. However, familiarity with advanced topics in differential equations might be helpful at times.

Course material

The course will follow the course notes. Students might also benefit by reading from the following books.

The Mathematical Theory of Optimal Processes
L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, E. F. Mishchenko
Gordon & Breach
1986.
This is the original reference (first published in Russian in 1961), and has in many ways not been excelled in its presentation of the Maximum Principle.

Foundations of Optimal Control Theory
E. B. Lee, L. Markus
John Wiley and Sons
1967.
This is an excellent text on optimal control in general. As concerns the Maximum Principle, it provides some alternative treatments to the original presentation by Pontryagin, et al. In particular, it isolates the relationship between the Maximum Principle and the boundary of the reachable set. We shall have a great deal to say about this in our presentation.

Optimal Control. An Introduction to the Theory and its Applications
M. Athans and P. L. Falb
McGraw-Hill
1966.
This text is written for an engineering audience. It therefore sacrifices some rigour to gain intuition. Nonetheless, it can be a valuable resource in the early stages of trying to understand the Maximum Principle. And it is also a good place to see how the Maximum Principle can be applied.