SNOPT优化工具箱

function [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun,varargin); % [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun,varargin); % This function solves the nonlinear optimization problem: % minimize: % F(ObjRow) + ObjAdd % subject to: % xlow <= x <= xupp % Flow <= F <= Fupp % where: % x is the column vector of initial values of the unknowns. % F is a vector of objective and constraint functions specified % in the m-file userfun. % ObjRow is the objective row of F (default ObjRow = 1). % userfun is a string containing the name of a user-defined m-file % that defines the elements of F and optionally, their % derivatives. % % Additional arguments allow the specification of more detailed % problem information. % % Calling sequence 1: % [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun) % % Calling sequence 2: % [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun, % ObjAdd, ObjRow) % Calling sequence 3: % [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun, % A, iAfun, jAvar, iGfun, jGvar) % Calling sequence 4: % [x,F,inform,xmul,Fmul] = snopt(x,xlow,xupp,Flow,Fupp,userfun, % ObjAdd, ObjRow, % A, iAfun, jAvar, iGfun, jGvar) % Description of the arguments: % x -- initial guess for x. % xlow, xupp -- upper and lower bounds on x. % Flow, Fupp -- upper and lower bounds on F. % userfun -- a string denoting name of a user-defined function % that computes the objective and constraint functions % and their corresponding derivatives. % WARNING: If arguments iAfun, jAvar, and A, are provided, % then it is crucial that the associated linear terms % are not included in the calculation of F in userfun. % **Example: % >> [f,G] = feval(userfun,x); % >> F = sparse(iAfun,jAvar,A)*x + f % >> DF = sparse(iAfun,jAvar,A) + sparse(iGfun,jGvar,G) % (where DF denotes F'). % G must be either a dense matrix, or a vector of derivatives % in the same order as the indices iGfun and jGvar, i.e., % if G is a vector, the Jacobian of f is given % by sparse(iGfun,jGvar,G). % % **More details on G: % The user maybe define, all, some, or none of the % entries of G. If the user does NOT intend to % supply ALL nonzero entries of G, it is imperative % that the proper derivative level is set prior to % a call to snopt(): % >> snseti("Derivative option",k); % where k = 0 or 1. Meaning: % 1 -- Default. All derivatives are provided. % 0 -- Some derivatives are not provided. % For the case k = 0 (ONLY), G may be returned as % an empty array []. % For the case k = 0, the user must denote % unknown NONZERO elements of G by NaN. % **Example: (vector case) % >> G = [1, NaN, 3, NaN, -5]'; % or (full matrix case) % >> G = [1, 0, NaN; 0 NaN 2; 0 0 3]; % ObjAdd -- Default 0. Constant added to F(ObjRow). % ObjRow -- Default 1. Denotes row of objective function in F. % A -- Constant elements in the Jacobian of F. % iAfun, jAvar -- Indices of A, corresponding to A. % iGfun, jGvar -- Indices of nonlinear elements in the Jacobian of F. % % More IMPORTANT details: % 1) The indices (iAfun,jAvar) must be DISJOINT from (iGfun,jGvar). % A nonzero element in F' must be either an element of G or an % element of A, but not the sum of the two. % % 2) If the user does not wish to provide iAfun, jAvar, iGfun, % jGvar, then snopt() will determine them by calling snJac(). % % WARNING: In this case, the derivative level will be set to zero % if constant elements exist. This is because the linear % elements have not yet been deleted from the definition of % userfun. Furthermore, if G is given in vector form, the % ordering of G may not necessarily correspond to (iGfun,jGvar) % computed by snJac(). m = length(Flow); n = length(x); xmul = zeros(n,1); xstate = zeros(n,1); Fmul = zeros(m,1); Fstate = zeros(m,1); ObjAdd = 0; ObjRow = 1; if nargin == 6 % Calling sequence 1 % The derivatives will be estimated by differences. % Call snJac to estimate the pattern of nonzeros for the Jacobian. [A,iAfun,jAvar,iGfun,jGvar] = snJac(userfun,x,xlow,xupp,m); % At this point, to supply derivatives, userfun must be modified. if (length(A) ~= 0) snset('Derivative option 0'); end elseif nargin == 8 % Calling sequence 2 % The derivatives will be estimated by differences. % Call snJac to estimate the pattern of nonzeros for the Jacobian. [A,iAfun,jAvar,iGfun,jGvar] = snJac(userfun,x,xlow,xupp,m); %At this point, to supply derivatives, userfun must be modified. if length(A) ~= 0 snset('Derivative option 0'); end ObjAdd = varargin{1}; ObjRow = varargin{2}; elseif nargin == 11 % Calling sequence 3 % The user is providing derivatives. A = varargin{1}; iAfun = varargin{2}; jAvar = varargin{3}; iGfun = varargin{4}; jGvar = varargin{5}; elseif ( nargin == 13 ) % Calling sequence 4 % The user is providing derivatives. ObjAdd = varargin{1}; ObjRow = varargin{2}; A = varargin{3}; iAfun = varargin{4}; jAvar = varargin{5}; iGfun = varargin{6}; jGvar = varargin{7}; end solveopt = 1; [x,F,xmul,Fmul,inform] = snoptcmex( solveopt, ... x,xlow,xupp,xmul,xstate, ... Flow,Fupp,Fmul,Fstate, ... ObjAdd,ObjRow,A,iAfun,jAvar, ... iGfun,jGvar,userfun );