New conditions for making a mathematical principle more helpful

Researchers identified specific problem classes where Pontryagin's Maximum Principle works effectively.

Pontryagin's Maximum Principle is helpful when it is normal. A solution is normal if it explicitly involves the cost function of the problem.

The work identifies classes of Lagrange optimal control problems without state constraints where the Principle is normal. General conditions are introduced to guarantee this normality.

One class involves an augmented reachable set with a convex lower-boundary function. The optimal terminal state must lie in the relative interior of the reachable state set.

The study looks at Lagrange steering problems. For these, normality is guaranteed if all eigenvalues of the state dynamics matrix AA are real or have small enough imaginary parts.

This property does not extend to arbitrary AA matrices. Simple 2×22 \times 2 examples show this difference from the finite-dimensional case.

Key facts

  • Normality occurs when a solution explicitly involves the problem's cost function.
  • Normality is guaranteed for certain steering problems if all eigenvalues of matrix AA are real.
  • The research identifies classes of problems where the augmented reachable set has a convex lower boundary.

All sections

Edition