Directory: | ./ |
---|---|
File: | src/path-optimization/spline-gradient-based/cost.hh |
Date: | 2024-08-10 11:29:48 |
Exec | Total | Coverage | |
---|---|---|---|
Lines: | 40 | 49 | 81.6% |
Branches: | 34 | 84 | 40.5% |
Line | Branch | Exec | Source |
---|---|---|---|
1 | // Copyright (c) 2017, Joseph Mirabel | ||
2 | // Authors: Joseph Mirabel (joseph.mirabel@laas.fr) | ||
3 | // | ||
4 | |||
5 | // Redistribution and use in source and binary forms, with or without | ||
6 | // modification, are permitted provided that the following conditions are | ||
7 | // met: | ||
8 | // | ||
9 | // 1. Redistributions of source code must retain the above copyright | ||
10 | // notice, this list of conditions and the following disclaimer. | ||
11 | // | ||
12 | // 2. Redistributions in binary form must reproduce the above copyright | ||
13 | // notice, this list of conditions and the following disclaimer in the | ||
14 | // documentation and/or other materials provided with the distribution. | ||
15 | // | ||
16 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | ||
17 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | ||
18 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | ||
19 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | ||
20 | // HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | ||
21 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | ||
22 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | ||
23 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | ||
24 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | ||
25 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | ||
26 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH | ||
27 | // DAMAGE. | ||
28 | |||
29 | #ifndef HPP_CORE_PATH_OPTIMIZATION_SPLINE_GRADIENT_BASED_COST_HH | ||
30 | #define HPP_CORE_PATH_OPTIMIZATION_SPLINE_GRADIENT_BASED_COST_HH | ||
31 | |||
32 | #include <hpp/core/path-optimization/cost.hh> | ||
33 | #include <hpp/core/path/spline.hh> | ||
34 | #include <hpp/util/debug.hh> | ||
35 | #include <hpp/util/exception-factory.hh> | ||
36 | |||
37 | namespace hpp { | ||
38 | namespace core { | ||
39 | namespace pathOptimization { | ||
40 | /// TODO | ||
41 | /// The derivative of the cost is wrong when for freeflyer and planar | ||
42 | /// joints. It lacks the derivative of the difference operator. The issue | ||
43 | /// is that it is not a quadratic cost anymore. | ||
44 | template <typename _Spline> | ||
45 | struct HPP_CORE_LOCAL L2NormSquaredOfDerivative { | ||
46 | typedef _Spline Spline; | ||
47 | typedef typename Spline::Ptr_t SplinePtr_t; | ||
48 | typedef std::vector<SplinePtr_t> Splines_t; | ||
49 | |||
50 | 30 | L2NormSquaredOfDerivative(const Splines_t& splines, size_type paramSize, | |
51 | size_type paramDerivativeSize, | ||
52 | size_type derivativeOrder) | ||
53 |
1/2✓ Branch 2 taken 15 times.
✗ Branch 3 not taken.
|
30 | : lambda_(splines.size()), |
54 | 30 | nSplines_(splines.size()), | |
55 | 30 | paramSize_(paramSize), | |
56 | 30 | paramDerivativeSize_(paramDerivativeSize), | |
57 | 30 | inputSize_(nSplines_ * Spline::NbCoeffs * paramSize), | |
58 | 30 | inputDerivativeSize_(nSplines_ * Spline::NbCoeffs * | |
59 | 30 | paramDerivativeSize), | |
60 | 30 | derivativeOrder_(derivativeOrder) { | |
61 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 15 times.
|
30 | assert(derivativeOrder_ > 0); |
62 | // Spline::NbPowerOfT = 2 * Spline::Order + 3 | ||
63 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 15 times.
|
30 | if (2 * derivativeOrder_ - 1 >= Spline::NbPowerOfT) { |
64 | ✗ | HPP_THROW(std::invalid_argument, | |
65 | "Cannot compute the squared norm of the " | ||
66 | << derivativeOrder_ | ||
67 | << "th order derivative with splines of order " | ||
68 | << Spline::Order); | ||
69 | } | ||
70 |
1/2✓ Branch 1 taken 15 times.
✗ Branch 2 not taken.
|
30 | lambda_.setOnes(); |
71 | 30 | } | |
72 | |||
73 | ✗ | void computeLambdasFromSplineLength(const Splines_t& splines) { | |
74 | ✗ | for (std::size_t i = 0; i < nSplines_; ++i) | |
75 | ✗ | lambda_[i] = splines[i]->squaredNormIntegral(derivativeOrder_); | |
76 | ✗ | value_type lMax = lambda_.maxCoeff(); | |
77 | // Make sure there is no too relatively small values in lambda_. | ||
78 | ✗ | lambda_ = (lambda_.array() > 1e-6 * lMax) | |
79 | ✗ | .select(lambda_.cwiseInverse(), 1e6 / lMax); | |
80 | } | ||
81 | |||
82 | 132 | void value(value_type& result, const Splines_t& splines) const { | |
83 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 66 times.
|
132 | assert(nSplines_ == splines.size()); |
84 | 132 | result = 0; | |
85 |
2/2✓ Branch 0 taken 228 times.
✓ Branch 1 taken 66 times.
|
588 | for (std::size_t i = 0; i < nSplines_; ++i) |
86 | 456 | result += lambda_[i] * splines[i]->squaredNormIntegral(derivativeOrder_); | |
87 | 132 | } | |
88 | |||
89 | void jacobian(vectorOut_t J, const Splines_t& splines) const { | ||
90 | assert(nSplines_ == splines.size()); | ||
91 | assert(J.size() == inputDerivativeSize_); | ||
92 | size_type col = 0; | ||
93 | size_type size = Spline::NbCoeffs * paramDerivativeSize_; | ||
94 | for (std::size_t i = 0; i < nSplines_; ++i) { | ||
95 | splines[i]->squaredNormIntegralDerivative(derivativeOrder_, | ||
96 | J.segment(col, size)); | ||
97 | J.segment(col, size) *= lambda_[i]; | ||
98 | col += size; | ||
99 | } | ||
100 | } | ||
101 | |||
102 | 30 | void hessian(matrixOut_t H, const Splines_t& splines) const { | |
103 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 15 times.
|
30 | assert(H.rows() == inputDerivativeSize_); |
104 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 15 times.
|
30 | assert(H.cols() == inputDerivativeSize_); |
105 |
1/2✓ Branch 1 taken 15 times.
✗ Branch 2 not taken.
|
30 | typename Spline::BasisFunctionIntegralMatrix_t Ic; |
106 | |||
107 |
1/2✓ Branch 1 taken 15 times.
✗ Branch 2 not taken.
|
30 | H.setZero(); |
108 | |||
109 |
2/2✓ Branch 0 taken 51 times.
✓ Branch 1 taken 15 times.
|
132 | for (std::size_t k = 0; k < nSplines_; ++k) { |
110 |
1/2✓ Branch 3 taken 51 times.
✗ Branch 4 not taken.
|
102 | splines[k]->squaredNormBasisFunctionIntegral(derivativeOrder_, Ic); |
111 |
1/2✓ Branch 1 taken 51 times.
✗ Branch 2 not taken.
|
102 | Ic *= 2; |
112 | 102 | const size_type shift = k * Spline::NbCoeffs * paramSize_; | |
113 |
2/2✓ Branch 0 taken 258 times.
✓ Branch 1 taken 51 times.
|
618 | for (size_type i = 0; i < Spline::NbCoeffs; ++i) { |
114 |
2/2✓ Branch 0 taken 1492 times.
✓ Branch 1 taken 258 times.
|
3500 | for (size_type j = 0; j < Spline::NbCoeffs; ++j) { |
115 | ✗ | H.block(shift + i * paramSize_, shift + j * paramSize_, paramSize_, | |
116 |
1/2✓ Branch 1 taken 1492 times.
✗ Branch 2 not taken.
|
2984 | paramSize_) |
117 | .diagonal() | ||
118 |
4/8✓ Branch 1 taken 1492 times.
✗ Branch 2 not taken.
✓ Branch 4 taken 1492 times.
✗ Branch 5 not taken.
✓ Branch 7 taken 1492 times.
✗ Branch 8 not taken.
✓ Branch 10 taken 1492 times.
✗ Branch 11 not taken.
|
2984 | .setConstant(Ic(i, j) * lambda_[k]); |
119 | } | ||
120 | } | ||
121 | |||
122 | #ifndef NDEBUG | ||
123 |
7/14✓ Branch 2 taken 51 times.
✗ Branch 3 not taken.
✓ Branch 7 taken 51 times.
✗ Branch 8 not taken.
✓ Branch 10 taken 51 times.
✗ Branch 11 not taken.
✓ Branch 13 taken 51 times.
✗ Branch 14 not taken.
✓ Branch 16 taken 51 times.
✗ Branch 17 not taken.
✓ Branch 19 taken 51 times.
✗ Branch 20 not taken.
✓ Branch 22 taken 51 times.
✗ Branch 23 not taken.
|
204 | value_type res1 = 0.5 * splines[k]->rowParameters().transpose() * |
124 | ✗ | H.block(shift, shift, Spline::NbCoeffs * paramSize_, | |
125 |
1/2✓ Branch 1 taken 51 times.
✗ Branch 2 not taken.
|
102 | Spline::NbCoeffs * paramSize_) * |
126 | 102 | splines[k]->rowParameters(); | |
127 | |||
128 | 102 | value_type res2 = | |
129 |
2/4✓ Branch 3 taken 51 times.
✗ Branch 4 not taken.
✓ Branch 6 taken 51 times.
✗ Branch 7 not taken.
|
102 | splines[k]->squaredNormIntegral(derivativeOrder_) * lambda_[k]; |
130 | |||
131 | 102 | value_type diff = res1 - res2; | |
132 | |||
133 | 102 | if (std::fabs(diff) > Eigen::NumTraits<value_type>::dummy_precision()) { | |
134 | hppDout(error, | ||
135 | "Hessian seems wrong for spline " | ||
136 | << k << ": " << res1 << " - " << res2 << " = " | ||
137 | << res1 - res2 << '\n' | ||
138 | << H.block(shift, shift, Spline::NbCoeffs * paramSize_, | ||
139 | Spline::NbCoeffs * paramSize_)); | ||
140 | } | ||
141 | #endif // NDEBUG | ||
142 | } | ||
143 | 30 | } | |
144 | |||
145 | vector_t lambda_; | ||
146 | const std::size_t nSplines_; | ||
147 | const size_type paramSize_, paramDerivativeSize_; | ||
148 | const size_type inputSize_, inputDerivativeSize_; | ||
149 | size_type derivativeOrder_; | ||
150 | }; | ||
151 | } // namespace pathOptimization | ||
152 | } // namespace core | ||
153 | } // namespace hpp | ||
154 | |||
155 | #endif // HPP_CORE_PATH_OPTIMIZATION_SPLINE_GRADIENT_BASED_COST_HH | ||
156 |