代码拉取完成，页面将自动刷新
<!DOCTYPE html>
<html>
<head>
<title>Underactuated Robotics: Optimization and
Mathematical Programming</title>
<meta name="Underactuated Robotics: Optimization and
Mathematical Programming" content="text/html; charset=utf-8;" />
<link rel="canonical" href="http://underactuated.mit.edu/optimization.html" />
<script src="https://hypothes.is/embed.js" async></script>
<script type="text/javascript" src="htmlbook/book.js"></script>
<script src="htmlbook/mathjax-config.js" defer></script>
<script type="text/javascript" id="MathJax-script" defer
src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-chtml.js">
</script>
<script>window.MathJax || document.write('<script type="text/javascript" src="htmlbook/MathJax/es5/tex-chtml.js" defer><\/script>')</script>
<link rel="stylesheet" href="htmlbook/highlight/styles/default.css">
<script src="htmlbook/highlight/highlight.pack.js"></script> <!-- http://highlightjs.readthedocs.io/en/latest/css-classes-reference.html#language-names-and-aliases -->
<script>hljs.initHighlightingOnLoad();</script>
<link rel="stylesheet" type="text/css" href="htmlbook/book.css" />
</head>
<body onload="loadChapter('underactuated');">
<div data-type="titlepage">
<header>
<h1><a href="index.html" style="text-decoration:none;">Underactuated Robotics</a></h1>
<p data-type="subtitle">Algorithms for Walking, Running, Swimming, Flying, and Manipulation</p>
<p style="font-size: 18px;"><a href="http://people.csail.mit.edu/russt/">Russ Tedrake</a></p>
<p style="font-size: 14px; text-align: right;">
© Russ Tedrake, 2020<br/>
<a href="tocite.html">How to cite these notes</a> |
<a target="_blank" href="https://docs.google.com/forms/d/e/1FAIpQLSesAhROfLRfexrRFebHWLtRpjhqtb8k_iEagWMkvc7xau08iQ/viewform?usp=sf_link">Send me your feedback</a><br/>
</p>
</header>
</div>
<p><b>Note:</b> These are working notes used for <a
href="http://underactuated.csail.mit.edu/Spring2020/">a course being taught
at MIT</a>. They will be updated throughout the Spring 2020 semester. <a
href="https://www.youtube.com/channel/UChfUOAhz7ynELF-s_1LPpWg">Lecture videos are available on YouTube</a>.</p>
<table style="width:100%;"><tr style="width:100%">
<td style="width:33%;text-align:left;"><a class="previous_chapter" href=multibody.html>Previous Chapter</a></td>
<td style="width:33%;text-align:center;"><a href=index.html>Table of contents</a></td>
<td style="width:33%;text-align:right;"><a class="next_chapter" href=playbook.html>Next Chapter</a></td>
</tr></table>
<!-- EVERYTHING ABOVE THIS LINE IS OVERWRITTEN BY THE INSTALL SCRIPT -->
<chapter class="appendix" style="counter-reset: chapter 2"><h1>Optimization and
Mathematical Programming</h1>
<p>The view of dynamics and controls taken in these notes builds heavily on
tools from optimization -- and our success in practice depends heavily on the
effective application of numerical optimization. There are many excellent
books on optimization, for example <elib>Nocedal06</elib> is an excellent
reference on smooth optimization and <elib>Boyd04a</elib> is an excellent
reference on convex optimization (which we use extensively). I will provide
more references for the specific optimization formulations below.</p>
<p>The intention of this chapter is, therefore, mainly to provide a launchpad
and to address some topics that are particularly relevant in robotics but
might be more hidden in the existing general treatments. You can get far very
quickly as a consumer of these tools with just a high-level understanding.
But, as with many things, sometimes the details matter and it becomes
important to understand what is going on under the hood. We can often
formulate an optimization problem in multiple ways that might be
mathematically equivalent, but perform very differently in practice.</p>
<section><h1>Optimization software</h1>
<p>Some of the algorithms from optimization are quite simple to implement
yourself; stochastic gradient descent is perhaps the classic example. Even
more of them are conceptually fairly simple, but for some of the algorithms,
the implementation details matter a great deal -- the difference between an
expert implementation and a novice implementation of the numerical recipe,
in terms of speed and/or robustness, can be dramatic. These packages often
use a wealth of techniques for numerically conditioning the problems, for
discarding trivially valid constraints, and for warm-starting optimization
between solves. Not all solvers support all types of objectives and
constraints, and even if we have two commercial solvers that both claim to
solve, e.g. quadratic programs, then they might perform differently on the
particular structure/conditioning of your problem. In some cases, we also
have nicely designed open-source alternatives to these commercial solvers
(often written by academics who are experts in optimization) -- sometimes
they compete well with the commercial solvers or even outperform them in
particular problem types.</p>
<p>As a result, there are also a handful of software packages that attempt
to provide a layer of abstraction between the formulation of a mathematical
program and the instantiation of that problem in each of the particular
solvers. Famous examples of this include <a
href="http://cvxr.com/cvx/">CVX</a> and <a
href="https://yalmip.github.io/">YALMIP</a> for MATLAB, and <a
href="https://www.juliaopt.org/JuMP.jl/stable/">JuMP</a> for Julia.
<drake></drake>'s MathematicalProgram classes provide a middle layer like
this for C++ and Python; its creation was motivated initially and
specifically by the need to support the optimization formulations we use in
this text.</p>
<p>We have a number of tutorials on mathematical programming in
<drake></drake>, starting with a general introduction <a
href="https://mybinder.org/v2/gh/RobotLocomotion/drake/nightly-release?filepath=tutorials/mathematical_program.ipynb">here</a>. Drake <a href="https://drake.mit.edu/doxygen_cxx/group__solvers.html">supports a number of custom, open-source, and commercial solvers</a> (and even some of the commercial solvers are free to academic users).</p>
</section>
<section><h1>General concepts</h1>
<p>The general formulation is ... </p>
<p>It's important to realize that even though this formulation is incredibly
general, it does have it's limits. As just one example, when write
optimizations to plan trajectories of a robot, in this formulation we
typically have to choose a-priori as particular number of decision variables
that will encode the solution. Although we can of course write algorithms
that change the number of variables and call the optimizer again, somehow I
feel that this "general" formulation fails to capture, for instance, the
type of mathematical programming that occurs in sample-based optimization
planning -- where the resulting solutions can be described by any finite
number of parameters, and the computation flexibly transitions amongst
them.</p>
<subsection><h1>Convex vs nonconvex optimization</h1>
<p>Local optima. Convex functions, convex constraints.</p>
<p>If an optimization problem is nonconvex, it does not necessarily mean
that the optimization is hard. There are many cases in deep learning where
we can reliably solve seemingly very high-dimensional and nonconvex
optimization problems. Our understanding of these phenomenon is evolving
rapidly, and I suspect anything I write here will shortly become outdated.
But the current thinking in supervised learning mostly revolves around the
idea that over-parameterization is key -- that many of these success
stories are happening in a regime where we actually have more decision
variables than we have data, and that the search space is actually dense
with solutions that can fit the data perfectly (the so-called
"interpolating solutions"), that not all global minima are equally robust,
and that the optimization algorithms are performing some form of implicit
regularization in order to pick a "good" interpolating solution.</p>
</subsection>
<subsection><h1>Constrained optimization with Lagrange multipliers</h1>
<p>Given the equality-constrained optimization problem $$\minimize_\bz
\ell(\bz) \quad \subjto \quad \bphi(\bz) = 0,$$ where $\bphi$ is a vector.
Define a vector $\lambda$ of Lagrange multipliers, the same size as
$\phi$, and the scalar Lagrangian function, $$L(\bz,{\bf \lambda}) =
\ell(\bz) + \lambda^T\phi(\bz).$$ A necessary condition for $\bz^*$ to be
an optimal value of the constrained optimization is that the gradients of
$L$ vanish with respect to both $\bz$ and $\lambda$: $$\pd{L}{\bz} = 0,
\quad \pd{L}{\lambda} = 0.$$ Note that $\pd{L}{\lambda} = \phi(\bz)$, so
requiring this to be zero is equivalent to requiring the constraints to be
satisfied.</p>
<example><h1>Optimization on the unit circle</h1>
<p>Consider the following optimization: $$\min_{x,y} x+y, \quad \subjto
\quad x^2 + y^2 = 1.$$ The level sets of $x+y$ are straight lines with
slope $-1$, and the constraint requires that the solution lives on the
unit circle.</p> <figure> <img width="80%"
src="figures/lagrange_multipliers_unit_circle.svg"/> </figure>
<p>Simply by inspection, we can determine that the optimal solution
should be $x=y=-\frac{\sqrt{2}}{2}.$ Let's make sure we can obtain the
same result using Lagrange multipliers. </p>
<p>Formulating $$L = x+y+\lambda(x^2+y^2-1),$$ we can take the
derivatives and solve
\begin{align*}
\pd{L}{x} = 1 + 2\lambda x = 0 \quad & \Rightarrow & \lambda = -\frac{1}{2x}, \\
\pd{L}{y} = 1 + 2\lambda y = 1 - \frac{y}{x} = 0 \quad & \Rightarrow & y = x,\\
\pd{L}{\lambda} = x^2+y^2-1 = 2x^2 -1 = 0 \quad & \Rightarrow & x = \pm \frac{1}{\sqrt{2}}.
\end{align*}
Given the two solutions which satisfy the necessary conditions, the
negative solution is clearly the minimizer of the objective.</p>
</example>
</subsection>
</section>
<section><h1>Convex optimization</h1>
<subsection><h1>Linear Programs/Quadratic Programs/Second-Order Cones</h1>
Example: Balancing force control on Atlas
</subsection>
<subsection><h1>Semidefinite Programming and Linear Matrix Inequalities</h1>
</subsection>
<subsection id=sums_of_squares><h1>Sums-of-squares optimization</h1>
<p>It turns out that in the same way that we can use SDP to search over
the positive quadratic equations, we can generalize this to search over
the positive polynomial equations. To be clear, for quadratic equations
we have \[ {\bf P}\succeq 0 \quad \Rightarrow \quad \bx^T {\bf P} \bx \ge
0, \quad \forall \bx. \] It turns out that we can generalize this to \[
{\bf P}\succeq 0 \quad \Rightarrow \quad {\bf m}^T(\bx) {\bf P} {\bf
m}(\bx) \ge 0, \quad \forall \bx, \] where ${\bf m}(\bx)$ is a vector of
polynomial equations, typically chosen as a vector of <em>monomials</em>
(polynomials with only one term). The set of positive polynomials
parameterized in this way is exactly the set of polynomials that can be
written as a <em>sum of squares</em><elib>Parrilo00</elib>. While it is
known that not all positive polynomials can be written in this form, much
is known about the gap. For our purposes this gap is very small (papers
have been written about trying to find polynomials which are uniformly
positive but not sums of squares); we should remember that it exists but
not worry about it too much for now.</p>
<p>Even better, there is quite a bit that is known about how to choose
the terms in ${\bf m}(\bx)$. For example, if you give me the polynomial
\[ p(x) = 2 - 4x + 5x^2 \] and ask if it is positive for all real $x$, I
can convince you that it is by producing the sums-of-squares
factorization \[ p(x) = 1 + (1-2x)^2 + x^2, \] and I know that I can
formulate the problem without needing any monomials of degree greater
than 1 (the square-root of the degree of $p$) in my monomial vector. In
practice, what this means for us is that people have authored
optimization front-ends which take a high-level specification with
constraints on the positivity of polynomials and they automatically
generate the SDP problem for you, without having to think about what
terms should appear in ${\bf m}(\bx)$ (c.f. <elib>Prajna04</elib>).
This class of optimization problems is called Sums-of-Squares (SOS)
optimization.</p>
<p>As it happens, the particular choice of ${\bf m}(\bx)$ can have a huge
impact on the numerics of the resulting semidefinite program (and on your
ability to solve it with commercial solvers). <drake></drake> implements
some particularly novel/advanced algorithms to make this work well
<elib>Permenter17</elib>.</p>
<p>We will write optimizations using sums-of-squares constraints as \[
p(\bx) \sos \] as shorthand for the constraint that $p(\bx) \ge 0$ for all
$\bx$, as demonstrated by finding a sums-of-squares decomposition.</p>
<p>This is surprisingly powerful. It allows us to use convex
optimization to solve what appear to be very non-convex optimization
problems.</p>
<example><h1>Global minimization via SOS</h1>
<p>Consider the following famous non-linear function of two variables
(called the "six-hump-camel": $$p(x) = 4x^2 + xy - 4y^2 - 2.1x^4 + 4y^4
+ \frac{1}{3}x^6.$$ This function has six local minima, two of them
being global minima <elib>Henrion09a</elib>.</p>
<div style="text-align:center"><img
src="figures/six_hump_camel.svg" width="90%"/></div>
<p>By formulating a simple sums-of-squares optimization, we can
actually find the minimum value of this function (technically, it is
only a lower bound, but in this case and many cases, it is surprisingly
tight) by writing: \begin{align*} \max_\lambda \ \ & \lambda \\
\text{s.t. } & p(x) - \lambda \text{ is sos.} \end{align*} Go ahead
and play with the code (most of the lines are only for plotting; the
actual optimization problem is nice and simple to formulate). </p>
<jupyter>examples/optimization.ipynb</jupyter>
<p>Note that this finds the minimum value, but does not actually
produce the $\bx$ value which mimimizes it. This is possible
<elib>Henrion09a</elib>, but it requires examining the dual of the
sums-of-squares solutions (which we don't need for the goals of this
chapter).</p>
</example>
<subsubsection><h1>Sums of squares on a Semi-Algebraic Set</h1>
<p>The S-procedure.</p>
</subsubsection>
<subsubsection><h1>Sums of squares optimization on an Algebraic Variety</h1>
<p>The S-procedure</p>
<p>Using the quotient ring</p>
<p>Quotient rings via sampling</p>
</subsubsection>
<subsubsection><h1>DSOS and SDSOS</h1>
</subsubsection>
</subsection>
<subsection><h1>Solution techniques</h1>
Interior point (Gurobi, Mosek, Sedumi, ...), First order methods
</subsection>
</section>
<section id="nonlinear"><h1>Nonlinear programming</h1>
<p>The generic formulation of a nonlinear optimization problem is \[
\min_z c(z) \quad \subjto \quad \bphi(z) \le 0, \] where $z$ is a
vector of <em>decision variables</em>, $c$ is a scalar <em>objective
function</em> and $\phi$ is a vector of <em>constraints</em>. Note
that, although we write $\phi \le 0$, this formulation captures
positivity constraints on the decision variables (simply multiply the
constraint by $-1$) and equality constraints (simply list both $\phi\le0$
and $-\phi\le0$) as well. </p>
<p>The picture that you should have in your head is a nonlinear, potentially
non-convex objective function defined over (multi-dimensional) $z$, with a
subset of possible $z$ values satisfying the constraints.</p>
<figure> <img width="80%"
src="figures/nonlinear_optimization_w_minima.svg"/>
<figcaption>One-dimensional cartoon of a nonlinear optimization problem. The
red dots represent local minima. The blue dot represents the optimal
solution.</figcaption> </figure>
<p>Note that minima can be the result of the objective function having zero
derivative <em>or</em> due to a sloped objective up against a
constraint.</p>
<p>Numerical methods for solving these optimization problems require an
initial guess, $\hat{z}$, and proceed by trying to move down the objective
function to a minima. Common approaches include <em>gradient descent</em>,
in which the gradient of the objective function is computed or estimated, or
second-order methods such as <em>sequential quadratic programming (SQP)</em>
which attempts to make a local quadratic approximation of the objective
function and local linear approximations of the constraints and solves a
quadratic program on each iteration to jump directly to the minimum of the
local approximation.</p>
<p>While not strictly required, these algorithms can often benefit a great
deal from having the gradients of the objective and constraints computed
explicitly; the alternative is to obtain them from numerical
differentiation. Beyond pure speed considerations, I strongly prefer to
compute the gradients explicitly because it can help avoid numerical
accuracy issues that can creep in with finite difference methods. The
desire to calculate these gradients will be a major theme in the discussion
below, and we have gone to great lengths to provide explicit gradients of
our provided functions and automatic differentiation of user-provided
functions in <drake></drake>.</p>
<p>When I started out, I was of the opinion that there is nothing difficult
about implementing gradient descent or even a second-order method, and I
wrote all of the solvers myself. I now realize that I was wrong. The
commercial solvers available for nonlinear programming are substantially
higher performance than anything I wrote myself, with a number of tricks,
subtleties, and parameter choices that can make a huge difference in
practice. Some of these solvers can exploit sparsity in the problem (e.g.,
if the constraints operate in a sparse way on the decision variables).
Nowadays, we make heaviest use of SNOPT <elib>Gill06</elib>, which now comes
bundled with the binary distributions of <drake></drake>, but also <a
href="http://drake.mit.edu/doxygen_cxx/group__solvers.html">support a large
suite of numerical solvers</a>. Note that while I do advocate using these
tools, you do not need to use them as a black box. In many cases you can
improve the optimization performance by understanding and selecting
non-default configuration parameters.</p>
<subsection><h1>Second-order methods (SQP /
Interior-Point)</h1></subsection>
<subsection><h1>First-order methods (SGD / ADMM) </h1>
<subsubsection id="penalty"><h1>Penalty methods</h1>
<p>Augmented Lagrangian</p>
</subsubsection>
<subsubsection><h1>Projected Gradient Descent</h1></subsubsection>
</subsection>
<subsection><h1>Zero-order methods (CMA)</h1></subsection>
<todo>My original list: Lagrange multipliers / KKT, Gradient descent, SQP (SNOPT, NLOPT, IPOPT), Global optimization</todo>
<subsection><h1>Example: Inverse Kinematics</h1>
</subsection>
</section> <!-- nonlinear optimization -->
<section><h1>Combinatorial optimization</h1>
<subsection><h1>Search, SAT, First order logic, SMT solvers, LP interpretation</h1></subsection>
<subsection><h1>Mixed-integer convex optimization</h1>
<p>An advanced, but very readable book on MIP <elib>Conforti14</elib>. Nice survey paper on MILP <elib>Vielma15</elib>.</p>
</subsection>
</section>
<section><h1>"Black-box" optimization</h1>
Derivative-free methods. Some allow noisy evaluations.
</section>
<todo>Polynomial root finding/homotopy, L1 optimization, QCQP, LCP/Variational inequalities, BMI, ..., warm-starting, ...</todo>
</chapter>
<!-- EVERYTHING BELOW THIS LINE IS OVERWRITTEN BY THE INSTALL SCRIPT -->
<table style="width:100%;"><tr style="width:100%">
<td style="width:33%;text-align:left;"><a class="previous_chapter" href=multibody.html>Previous Chapter</a></td>
<td style="width:33%;text-align:center;"><a href=index.html>Table of contents</a></td>
<td style="width:33%;text-align:right;"><a class="next_chapter" href=playbook.html>Next Chapter</a></td>
</tr></table>
<div id="footer">
<hr>
<table style="width:100%;">
<tr><td><em>Underactuated Robotics</em></td><td align="right">© Russ
Tedrake, 2020</td></tr>
</table>
</div>
</body>
</html>
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。
马建仓 AI 助手