Viscocity Solutions of Hamilton-Jacobi Equations and Optimal Control Problems. Alberto Bressan, S.I.S.S.A, Trieste, Italy.
A linear pendulum problem controlled by an external force.
Find u over t in [0; 20 ] to maximize:
| J = x1(tf) |
subject to:
| = x2 |
| = u−x1 |
| x(t0) = [0 0] |
| |u| <= 1 |
Reference: [8]
toms t
t_f = 20;
p = tomPhase('p', t, 0, t_f, 60);
setPhase(p);
tomStates x1 x2
tomControls u
% Initial guess
x0 = {icollocate({x1 == 0; x2 == 0})
collocate(u == 0)};
% Box constraints and bounds
cb = {-1 <= collocate(u) <= 1
initial(x1 == 0)
initial(x2 == 0)};
% ODEs and path constraints
ceq = collocate({dot(x1) == x2
dot(x2) == u-x1});
% Objective
objective = -final(x1);
options = struct;
options.name = 'Linear Pendulum';
solution = ezsolve(objective, {cb, ceq}, x0, options);
t = subs(collocate(t),solution);
x1 = subs(collocate(x1),solution);
x2 = subs(collocate(x2),solution);
u = subs(collocate(u),solution);
Problem type appears to be: lp
Starting numeric solver
===== * * * =================================================================== * * *
TOMLAB - Tomlab Optimization Inc. Development license 999001. Valid to 2011-02-05
=====================================================================================
Problem: --- 1: Linear Pendulum f_k -12.612222977985938000
sum(|constr|) 0.000000000003687556
f(x_k) + sum(|constr|) -12.612222977982251000
f(x_0) 0.000000000000000000
Solver: CPLEX. EXIT=0. INFORM=1.
CPLEX Dual Simplex LP solver
Optimal solution found
FuncEv 206 Iter 206
CPU time: 0.031250 sec. Elapsed time: 0.031000 sec.
subplot(3,1,1)
plot(t,x1,'*-',t,x2,'*-');
legend('x1','x2');
title('Linear Pendulum state variables');
subplot(3,1,2)
plot(t,u,'+-');
legend('u');
title('Linear Pendulum control');
subplot(3,1,3)
plot(t,sign(sin(t_f-t)),'*-');
legend('Known u');
title('Linear Pendulum known solution');