model predictive control

%定义一个类，下面的是main函数的参数 classdef MPC properties A; B; Q; R; N; Ts; x_min; x_max; u_min; u_max; end methods %这部分可以看官网介绍，没什么就是传参调用 function obj = MPC(A, B, Q, R, N, Ts, x_min, x_max, u_min, u_max) obj.A = A; obj.B = B; obj.Q = Q; obj.R = R; obj.N = N; obj.Ts = Ts; obj.x_min = x_min; obj.x_max = x_max; obj.u_min = u_min; obj.u_max = u_max; end % 求解最优控制策略,只取最优输入量，下面一个类似的是取10个预测量 function u_opt = optimize(obj, x0) n_states = size(obj.A, 1); n_controls = size(obj.B, 2); opti = casadi.Opti(); X = opti.variable(n_states, obj.N+1); U = opti.variable(n_controls, obj.N); opti.subject_to(X(:,1) == x0); for k = 1:obj.N opti.subject_to(X(:,k+1) == obj.A*X(:,k) + obj.B*U(:,k)); end cost = 0; for k = 1:obj.N cost = cost + X(:,k)'*obj.Q*X(:,k) + U(:,k)'*obj.R*U(:,k); end opti.subject_to(obj.x_min <= X <= obj.x_max); opti.subject_to(obj.u_min <= U <= obj.u_max); % 优化代价函数 opti.minimize(cost); %结构体 ops = struct(); ops.ipopt.print_level = 0; ops.print_time = 0; ops.ipopt.tol = 1e-3; opti.solver('ipopt', ops); %求解 sol = opti.solve(); u_ten=sol.value(U(:,1:10)); u_opt = sol.value(U(:,1)); end function u_ten = optimizel(obj, x0) n_states = size(obj.A, 1); n_controls = size(obj.B, 2); opti = casadi.Opti(); X = opti.variable(n_states, obj.N+1); U = opti.variable(n_controls, obj.N); opti.subject_to(X(:,1) == x0); for k = 1:obj.N opti.subject_to(X(:,k+1) == obj.A*X(:,k) + obj.B*U(:,k)); end cost = 0; for k = 1:obj.N cost = cost + X(:,k)'*obj.Q*X(:,k) + U(:,k)'*obj.R*U(:,k); end opti.subject_to(obj.x_min <= X <= obj.x_max); opti.subject_to(obj.u_min <= U <= obj.u_max); opti.minimize(cost); ops = struct(); ops.ipopt.print_level = 0; ops.print_time = 0; ops.ipopt.tol = 1e-3; opti.solver('ipopt', ops); sol = opti.solve(); u_ten=sol.value(U(:,1:10)); end end end