1 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
2 |
|
|
// BSD 3-Clause License |
3 |
|
|
// |
4 |
|
|
// Copyright (C) 2019-2020, University of Edinburgh |
5 |
|
|
// Copyright note valid unless otherwise stated in individual files. |
6 |
|
|
// All rights reserved. |
7 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
8 |
|
|
|
9 |
|
|
#include <iostream> |
10 |
|
|
|
11 |
|
|
#include "crocoddyl/core/solvers/box-qp.hpp" |
12 |
|
|
#include "crocoddyl/core/utils/timer.hpp" |
13 |
|
|
|
14 |
|
|
int main(int argc, char* argv[]) { |
15 |
|
|
unsigned int NX = 36; // dimension of the decision vector |
16 |
|
|
unsigned int T = 5e3; // number of trials |
17 |
|
|
if (argc > 1) { |
18 |
|
|
T = atoi(argv[1]); |
19 |
|
|
} |
20 |
|
|
|
21 |
|
|
// Solving the bounded QP problem |
22 |
|
|
crocoddyl::BoxQP boxqp(NX); |
23 |
|
|
Eigen::ArrayXd duration(T); |
24 |
|
|
for (unsigned int i = 0; i < T; ++i) { |
25 |
|
|
// Creating a new random problem |
26 |
|
|
Eigen::MatrixXd H = Eigen::MatrixXd::Random(NX, NX); |
27 |
|
|
Eigen::MatrixXd hessian = H.transpose() * H; |
28 |
|
|
hessian = 0.5 * (hessian + hessian.transpose()).eval(); |
29 |
|
|
Eigen::VectorXd gradient = Eigen::VectorXd::Random(NX); |
30 |
|
|
Eigen::VectorXd lb = Eigen::VectorXd::Zero(NX); |
31 |
|
|
Eigen::VectorXd ub = Eigen::VectorXd::Ones(NX); |
32 |
|
|
Eigen::VectorXd xinit(NX); |
33 |
|
|
|
34 |
|
|
crocoddyl::Timer timer; |
35 |
|
|
boxqp.solve(hessian, gradient, lb, ub, xinit); |
36 |
|
|
duration[i] = timer.get_duration(); |
37 |
|
|
} |
38 |
|
|
|
39 |
|
|
double avrg_duration = duration.sum() / T; |
40 |
|
|
double min_duration = duration.minCoeff(); |
41 |
|
|
double max_duration = duration.maxCoeff(); |
42 |
|
|
std::cout << " BoxQP.solve (36) [ms]: " << avrg_duration << " (" |
43 |
|
|
<< min_duration << "-" << max_duration << ")" << std::endl; |
44 |
|
|
|
45 |
|
|
NX = 76; |
46 |
|
|
boxqp.set_nx(NX); |
47 |
|
|
for (unsigned int i = 0; i < T; ++i) { |
48 |
|
|
// Creating a new random problem |
49 |
|
|
Eigen::MatrixXd H = Eigen::MatrixXd::Random(NX, NX); |
50 |
|
|
Eigen::MatrixXd hessian = H.transpose() * H; |
51 |
|
|
hessian = 0.5 * (hessian + hessian.transpose()).eval(); |
52 |
|
|
Eigen::VectorXd gradient = Eigen::VectorXd::Random(NX); |
53 |
|
|
Eigen::VectorXd lb = Eigen::VectorXd::Zero(NX); |
54 |
|
|
Eigen::VectorXd ub = Eigen::VectorXd::Ones(NX); |
55 |
|
|
Eigen::VectorXd xinit(NX); |
56 |
|
|
|
57 |
|
|
crocoddyl::Timer timer; |
58 |
|
|
boxqp.solve(hessian, gradient, lb, ub, xinit); |
59 |
|
|
duration[i] = timer.get_duration(); |
60 |
|
|
} |
61 |
|
|
|
62 |
|
|
avrg_duration = duration.sum() / T; |
63 |
|
|
min_duration = duration.minCoeff(); |
64 |
|
|
max_duration = duration.maxCoeff(); |
65 |
|
|
std::cout << " BoxQP.solve (76) [ms]: " << avrg_duration << " (" |
66 |
|
|
<< min_duration << "-" << max_duration << ")" << std::endl; |
67 |
|
|
} |