1 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
2 |
|
|
// BSD 3-Clause License |
3 |
|
|
// |
4 |
|
|
// Copyright (C) 2019-2023, LAAS-CNRS, University of Edinburgh |
5 |
|
|
// Heriot-Watt University |
6 |
|
|
// Copyright note valid unless otherwise stated in individual files. |
7 |
|
|
// All rights reserved. |
8 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
9 |
|
|
|
10 |
|
|
#include "crocoddyl/core/solvers/fddp.hpp" |
11 |
|
|
#include "crocoddyl/core/utils/callbacks.hpp" |
12 |
|
|
#include "crocoddyl/core/utils/timer.hpp" |
13 |
|
|
#include "factory/arm.hpp" |
14 |
|
|
|
15 |
|
|
int main(int argc, char* argv[]) { |
16 |
|
|
bool CALLBACKS = false; |
17 |
|
|
unsigned int N = 100; // number of nodes |
18 |
|
|
unsigned int T = 5e3; // number of trials |
19 |
|
|
unsigned int MAXITER = 1; |
20 |
|
|
if (argc > 1) { |
21 |
|
|
T = atoi(argv[1]); |
22 |
|
|
} |
23 |
|
|
|
24 |
|
|
// Building the running and terminal models |
25 |
|
|
boost::shared_ptr<crocoddyl::ActionModelAbstract> runningModel, terminalModel; |
26 |
|
|
crocoddyl::benchmark::build_arm_action_models(runningModel, terminalModel); |
27 |
|
|
|
28 |
|
|
// Get the initial state |
29 |
|
|
boost::shared_ptr<crocoddyl::StateMultibody> state = |
30 |
|
|
boost::static_pointer_cast<crocoddyl::StateMultibody>( |
31 |
|
|
runningModel->get_state()); |
32 |
|
|
std::cout << "NQ: " << state->get_nq() << std::endl; |
33 |
|
|
std::cout << "Number of nodes: " << N << std::endl; |
34 |
|
|
Eigen::VectorXd q0 = |
35 |
|
|
state->get_pinocchio()->referenceConfigurations["arm_up"]; |
36 |
|
|
Eigen::VectorXd x0(state->get_nx()); |
37 |
|
|
x0 << q0, Eigen::VectorXd::Random(state->get_nv()); |
38 |
|
|
|
39 |
|
|
// For this optimal control problem, we define 100 knots (or running action |
40 |
|
|
// models) plus a terminal knot |
41 |
|
|
std::vector<boost::shared_ptr<crocoddyl::ActionModelAbstract> > runningModels( |
42 |
|
|
N, runningModel); |
43 |
|
|
boost::shared_ptr<crocoddyl::ShootingProblem> problem = |
44 |
|
|
boost::make_shared<crocoddyl::ShootingProblem>(x0, runningModels, |
45 |
|
|
terminalModel); |
46 |
|
|
std::vector<Eigen::VectorXd> xs(N + 1, x0); |
47 |
|
|
std::vector<Eigen::VectorXd> us( |
48 |
|
|
N, Eigen::VectorXd::Zero(runningModel->get_nu())); |
49 |
|
|
for (unsigned int i = 0; i < N; ++i) { |
50 |
|
|
const boost::shared_ptr<crocoddyl::ActionModelAbstract>& model = |
51 |
|
|
problem->get_runningModels()[i]; |
52 |
|
|
const boost::shared_ptr<crocoddyl::ActionDataAbstract>& data = |
53 |
|
|
problem->get_runningDatas()[i]; |
54 |
|
|
model->quasiStatic(data, us[i], x0); |
55 |
|
|
} |
56 |
|
|
|
57 |
|
|
// Formulating the optimal control problem |
58 |
|
|
crocoddyl::SolverFDDP solver(problem); |
59 |
|
|
if (CALLBACKS) { |
60 |
|
|
std::vector<boost::shared_ptr<crocoddyl::CallbackAbstract> > cbs; |
61 |
|
|
cbs.push_back(boost::make_shared<crocoddyl::CallbackVerbose>()); |
62 |
|
|
solver.setCallbacks(cbs); |
63 |
|
|
} |
64 |
|
|
|
65 |
|
|
// Solving the optimal control problem |
66 |
|
|
Eigen::ArrayXd duration(T); |
67 |
|
|
for (unsigned int i = 0; i < T; ++i) { |
68 |
|
|
crocoddyl::Timer timer; |
69 |
|
|
solver.solve(xs, us, MAXITER, false, 0.1); |
70 |
|
|
duration[i] = timer.get_duration(); |
71 |
|
|
} |
72 |
|
|
|
73 |
|
|
double avrg_duration = duration.sum() / T; |
74 |
|
|
double min_duration = duration.minCoeff(); |
75 |
|
|
double max_duration = duration.maxCoeff(); |
76 |
|
|
std::cout << " FDDP.solve [ms]: " << avrg_duration << " (" << min_duration |
77 |
|
|
<< "-" << max_duration << ")" << std::endl; |
78 |
|
|
|
79 |
|
|
// Running calc |
80 |
|
|
for (unsigned int i = 0; i < T; ++i) { |
81 |
|
|
crocoddyl::Timer timer; |
82 |
|
|
problem->calc(xs, us); |
83 |
|
|
duration[i] = timer.get_duration(); |
84 |
|
|
} |
85 |
|
|
|
86 |
|
|
avrg_duration = duration.sum() / T; |
87 |
|
|
min_duration = duration.minCoeff(); |
88 |
|
|
max_duration = duration.maxCoeff(); |
89 |
|
|
std::cout << " ShootingProblem.calc [ms]: " << avrg_duration << " (" |
90 |
|
|
<< min_duration << "-" << max_duration << ")" << std::endl; |
91 |
|
|
|
92 |
|
|
// Running calcDiff |
93 |
|
|
for (unsigned int i = 0; i < T; ++i) { |
94 |
|
|
crocoddyl::Timer timer; |
95 |
|
|
problem->calcDiff(xs, us); |
96 |
|
|
duration[i] = timer.get_duration(); |
97 |
|
|
} |
98 |
|
|
|
99 |
|
|
avrg_duration = duration.sum() / T; |
100 |
|
|
min_duration = duration.minCoeff(); |
101 |
|
|
max_duration = duration.maxCoeff(); |
102 |
|
|
std::cout << " ShootingProblem.calcDiff [ms]: " << avrg_duration << " (" |
103 |
|
|
<< min_duration << "-" << max_duration << ")" << std::endl; |
104 |
|
|
} |