The work of Peter Crouch the control theorist

  • Slides: 20
Download presentation
The work of Peter Crouch the control theorist* Conference on Decision and Control December

The work of Peter Crouch the control theorist* Conference on Decision and Control December 11, 2011 Canonical Geometrical Control Problems: New and Old Roger Brockett Engineering and Applied Sciences Harvard University *Not to be confused with the bad-boy English footballer of Tottenham Hotspur, Stoke City, Abigail Clancy, etc.

Peter Crouch: The reason we are here! Some of my early interactions: The London

Peter Crouch: The reason we are here! Some of my early interactions: The London NATO meeting– September, 1973 Student at Harvard, 1974 -1977: Thesis: “Dynamical Realizations of Finite Volterra Series” It showed that the natural state space for a finite Volterra series is diffeomorphic to Rn Cohort included P. S. Krishnaprassad and Joseph Ja’ Ja” Sabbatical at Harvard in 1982

Peter Crouch at the Center: From the Web

Peter Crouch at the Center: From the Web

Some Lie Theoretic, Least Squares, State Transfer Problems involving Z 2 Graded Lie Algebras

Some Lie Theoretic, Least Squares, State Transfer Problems involving Z 2 Graded Lie Algebras

The first two have finite Volterra series

The first two have finite Volterra series

Recall

Recall

What about regulator versions of these systems?

What about regulator versions of these systems?

What it Approximates

What it Approximates

Our Quadratic Regulator Problem

Our Quadratic Regulator Problem

The Euler-Lagrange Equations We need to factor the linear operator into a stable and

The Euler-Lagrange Equations We need to factor the linear operator into a stable and unstable factors. The value of x(0) is given. Its derivative is to be determined so as to put x on the right submanifold

Factoring the Euler-Lagrange Equation This is from the zeroth order term. This is from

Factoring the Euler-Lagrange Equation This is from the zeroth order term. This is from the first order term. Formula for Z

Relating Properties of x and Z through Q It is important that we are

Relating Properties of x and Z through Q It is important that we are now dealing with initial values Theta and Q are functions of x(0) and Z(0). -

Here we first define the optimal trajectory using initial conditions giving an open loop

Here we first define the optimal trajectory using initial conditions giving an open loop control. Actually it is true at all times and states! If considered as a “gain” is homogeneous of degree zero From the perspective of achieving the correct homogeneity, this is quite remarkable, even miraculous.

An Example These solutions are stable for all $a$ and generate a Z displacement.

An Example These solutions are stable for all $a$ and generate a Z displacement.

A Further Elaboration

A Further Elaboration

A Further Elaboration As x(0) approaches 0 the cost is upper bounded by the

A Further Elaboration As x(0) approaches 0 the cost is upper bounded by the cost of the u-only optimal trajectory. However, this cost is not differentiable on the “Z axis”.

As for the Cost---

As for the Cost---

This is not a dead end—Many more possibilities

This is not a dead end—Many more possibilities

Peter--Congratulations on a distinguished career based on talent, hard work, discipline, service to the

Peter--Congratulations on a distinguished career based on talent, hard work, discipline, service to the community.