DOLFINx 0.9.0
DOLFINx C++ interface
Loading...
Searching...
No Matches
gjk.h
1// Copyright (C) 2020 Chris Richardson
2//
3// This file is part of DOLFINx (https://www.fenicsproject.org)
4//
5// SPDX-License-Identifier: LGPL-3.0-or-later
6
7#pragma once
8
9#include <algorithm>
10#include <array>
11#include <concepts>
12#include <dolfinx/common/math.h>
13#include <limits>
14#include <numeric>
15#include <span>
16#include <utility>
17#include <vector>
18
19namespace dolfinx::geometry
20{
21
22namespace impl_gjk
23{
24
28template <std::floating_point T>
29std::pair<std::vector<T>, std::array<T, 3>>
30nearest_simplex(std::span<const T> s)
31{
32 assert(s.size() % 3 == 0);
33 const std::size_t s_rows = s.size() / 3;
34 switch (s_rows)
35 {
36 case 2:
37 {
38 // Compute lm = dot(s0, ds / |ds|)
39 auto s0 = s.template subspan<0, 3>();
40 auto s1 = s.template subspan<3, 3>();
41 std::array ds = {s1[0] - s0[0], s1[1] - s0[1], s1[2] - s0[2]};
42 const T lm = -(s0[0] * ds[0] + s0[1] * ds[1] + s0[2] * ds[2])
43 / (ds[0] * ds[0] + ds[1] * ds[1] + ds[2] * ds[2]);
44 if (lm >= 0.0 and lm <= 1.0)
45 {
46 // The origin is between A and B
47 // v = s0 + lm * (s1 - s0);
48 std::array v
49 = {s0[0] + lm * ds[0], s0[1] + lm * ds[1], s0[2] + lm * ds[2]};
50 return {std::vector<T>(s.begin(), s.end()), v};
51 }
52
53 if (lm < 0.0)
54 return {std::vector<T>(s0.begin(), s0.end()), {s0[0], s0[1], s0[2]}};
55 else
56 return {std::vector<T>(s1.begin(), s1.end()), {s1[0], s1[1], s1[2]}};
57 }
58 case 3:
59 {
60 auto a = s.template subspan<0, 3>();
61 auto b = s.template subspan<3, 3>();
62 auto c = s.template subspan<6, 3>();
63 auto length = [](auto& x, auto& y)
64 {
65 return std::transform_reduce(x.begin(), x.end(), y.begin(), 0.0,
66 std::plus{}, [](auto x, auto y)
67 { return (x - y) * (x - y); });
68 };
69 const T ab2 = length(a, b);
70 const T ac2 = length(a, c);
71 const T bc2 = length(b, c);
72
73 // Helper to compute dot(x, x - y)
74 auto helper = [](auto& x, auto& y)
75 {
76 return std::transform_reduce(x.begin(), x.end(), y.begin(), 0.0,
77 std::plus{},
78 [](auto x, auto y) { return x * (x - y); });
79 };
80 const std::array lm
81 = {helper(a, b) / ab2, helper(a, c) / ac2, helper(b, c) / bc2};
82
83 T caba = 0;
84 for (std::size_t i = 0; i < 3; ++i)
85 caba += (c[i] - a[i]) * (b[i] - a[i]);
86
87 // Calculate triangle ABC
88 const T c2 = 1 - caba * caba / (ab2 * ac2);
89 const T lbb = (lm[0] - lm[1] * caba / ab2) / c2;
90 const T lcc = (lm[1] - lm[0] * caba / ac2) / c2;
91
92 // Intersects triangle
93 if (lbb >= 0.0 and lcc >= 0.0 and (lbb + lcc) <= 1.0)
94 {
95 // Calculate intersection more accurately
96 // v = (c - a) x (b - a)
97 std::array dx0 = {c[0] - a[0], c[1] - a[1], c[2] - a[2]};
98 std::array dx1 = {b[0] - a[0], b[1] - a[1], b[2] - a[2]};
99 std::array v = math::cross(dx0, dx1);
100
101 // Barycentre of triangle
102 std::array p = {(a[0] + b[0] + c[0]) / 3, (a[1] + b[1] + c[1]) / 3,
103 (a[2] + b[2] + c[2]) / 3};
104
105 T sum = v[0] * p[0] + v[1] * p[1] + v[2] * p[2];
106 T vnorm2 = v[0] * v[0] + v[1] * v[1] + v[2] * v[2];
107 for (std::size_t i = 0; i < 3; ++i)
108 v[i] *= sum / vnorm2;
109
110 return {std::vector(s.begin(), s.end()), v};
111 }
112
113 // Get closest point
114 std::size_t pos = 0;
115 {
116 T norm0 = std::numeric_limits<T>::max();
117 for (std::size_t i = 0; i < s.size(); i += 3)
118 {
119 std::span<const T, 3> p(s.data() + i, 3);
120 T norm = p[0] * p[0] + p[1] * p[1] + p[2] * p[2];
121 if (norm < norm0)
122 {
123 pos = i / 3;
124 norm0 = norm;
125 }
126 }
127 }
128
129 std::array vmin = {s[3 * pos], s[3 * pos + 1], s[3 * pos + 2]};
130 T qmin = 0;
131 for (std::size_t k = 0; k < 3; ++k)
132 qmin += vmin[k] * vmin[k];
133
134 std::vector smin = {vmin[0], vmin[1], vmin[2]};
135
136 // Check if edges are closer
137 constexpr int f[3][2] = {{0, 1}, {0, 2}, {1, 2}};
138 for (std::size_t i = 0; i < s_rows; ++i)
139 {
140 auto s0 = s.subspan(3 * f[i][0], 3);
141 auto s1 = s.subspan(3 * f[i][1], 3);
142 if (lm[i] > 0 and lm[i] < 1)
143 {
144 std::array<T, 3> v;
145 for (std::size_t k = 0; k < 3; ++k)
146 v[k] = s0[k] + lm[i] * (s1[k] - s0[k]);
147 T qnorm = 0;
148 for (std::size_t k = 0; k < 3; ++k)
149 qnorm += v[k] * v[k];
150 if (qnorm < qmin)
151 {
152 std::ranges::copy(v, vmin.begin());
153 qmin = qnorm;
154 smin.resize(2 * 3);
155 std::span<T, 3> smin0(smin.data(), 3);
156 std::ranges::copy(s0, smin0.begin());
157 std::span<T, 3> smin1(smin.data() + 3, 3);
158 std::ranges::copy(s1, smin1.begin());
159 }
160 }
161 }
162 return {std::move(smin), vmin};
163 }
164 case 4:
165 {
166 auto s0 = s.template subspan<0, 3>();
167 auto s1 = s.template subspan<3, 3>();
168 auto s2 = s.template subspan<6, 3>();
169 auto s3 = s.template subspan<9, 3>();
170 auto W1 = math::cross(s0, s1);
171 auto W2 = math::cross(s2, s3);
172
173 std::array<T, 4> B;
174 B[0] = std::transform_reduce(s2.begin(), s2.end(), W1.begin(), 0.0);
175 B[1] = -std::transform_reduce(s3.begin(), s3.end(), W1.begin(), 0.0);
176 B[2] = std::transform_reduce(s0.begin(), s0.end(), W2.begin(), 0.0);
177 B[3] = -std::transform_reduce(s1.begin(), s1.end(), W2.begin(), 0.0);
178
179 const bool signDetM = std::signbit(std::reduce(B.begin(), B.end(), 0.0));
180 std::array<bool, 4> f_inside;
181 for (int i = 0; i < 4; ++i)
182 f_inside[i] = (std::signbit(B[i]) == signDetM);
183
184 if (f_inside[1] and f_inside[2] and f_inside[3])
185 {
186 if (f_inside[0]) // The origin is inside the tetrahedron
187 return {std::vector<T>(s.begin(), s.end()), {0, 0, 0}};
188 else // The origin projection P faces BCD
189 return nearest_simplex<T>(s.template subspan<0, 3 * 3>());
190 }
191
192 // Test ACD, ABD and/or ABC
193 std::vector<T> smin;
194 std::array<T, 3> vmin = {0, 0, 0};
195 constexpr int facets[3][3] = {{0, 1, 3}, {0, 2, 3}, {1, 2, 3}};
196 T qmin = std::numeric_limits<T>::max();
197 std::vector<T> M(9);
198 for (int i = 0; i < 3; ++i)
199 {
200 if (f_inside[i + 1] == false)
201 {
202 std::copy_n(std::next(s.begin(), 3 * facets[i][0]), 3, M.begin());
203 std::copy_n(std::next(s.begin(), 3 * facets[i][1]), 3,
204 std::next(M.begin(), 3));
205 std::copy_n(std::next(s.begin(), 3 * facets[i][2]), 3,
206 std::next(M.begin(), 6));
207
208 const auto [snew, v] = nearest_simplex<T>(M);
209 T q = std::transform_reduce(v.begin(), v.end(), v.begin(), 0);
210 if (q < qmin)
211 {
212 qmin = q;
213 vmin = v;
214 smin = snew;
215 }
216 }
217 }
218 return {smin, vmin};
219 }
220 default:
221 throw std::runtime_error("Number of rows defining simplex not supported.");
222 }
223}
224
226template <std::floating_point T>
227std::array<T, 3> support(std::span<const T> bd, std::array<T, 3> v)
228{
229 int i = 0;
230 T qmax = bd[0] * v[0] + bd[1] * v[1] + bd[2] * v[2];
231 for (std::size_t m = 1; m < bd.size() / 3; ++m)
232 {
233 T q = bd[3 * m] * v[0] + bd[3 * m + 1] * v[1] + bd[3 * m + 2] * v[2];
234 if (q > qmax)
235 {
236 qmax = q;
237 i = m;
238 }
239 }
240
241 return {bd[3 * i], bd[3 * i + 1], bd[3 * i + 2]};
242}
243} // namespace impl_gjk
244
255template <std::floating_point T>
256std::array<T, 3> compute_distance_gjk(std::span<const T> p,
257 std::span<const T> q)
258{
259 assert(p.size() % 3 == 0);
260 assert(q.size() % 3 == 0);
261
262 constexpr int maxk = 15; // Maximum number of iterations of the GJK algorithm
263
264 // Tolerance
265 constexpr T eps = 1.0e4 * std::numeric_limits<T>::epsilon();
266
267 // Initialise vector and simplex
268 std::array<T, 3> v = {p[0] - q[0], p[1] - q[1], p[2] - q[2]};
269 std::vector<T> s = {v[0], v[1], v[2]};
270
271 // Begin GJK iteration
272 int k;
273 for (k = 0; k < maxk; ++k)
274 {
275 // Support function
276 std::array w1 = impl_gjk::support(p, {-v[0], -v[1], -v[2]});
277 std::array w0 = impl_gjk::support(q, {v[0], v[1], v[2]});
278 const std::array w = {w1[0] - w0[0], w1[1] - w0[1], w1[2] - w0[2]};
279
280 // Break if any existing points are the same as w
281 assert(s.size() % 3 == 0);
282 std::size_t m;
283 for (m = 0; m < s.size() / 3; ++m)
284 {
285 auto it = std::next(s.begin(), 3 * m);
286 if (std::equal(it, std::next(it, 3), w.begin(), w.end()))
287 break;
288 }
289
290 if (m != s.size() / 3)
291 break;
292
293 // 1st exit condition (v - w).v = 0
294 const T vnorm2 = v[0] * v[0] + v[1] * v[1] + v[2] * v[2];
295 const T vw = vnorm2 - (v[0] * w[0] + v[1] * w[1] + v[2] * w[2]);
296 if (vw < (eps * vnorm2) or vw < eps)
297 break;
298
299 // Add new vertex to simplex
300 s.insert(s.end(), w.begin(), w.end());
301
302 // Find nearest subset of simplex
303 auto [snew, vnew] = impl_gjk::nearest_simplex<T>(s);
304 s.assign(snew.data(), snew.data() + snew.size());
305 v = {vnew[0], vnew[1], vnew[2]};
306
307 // 2nd exit condition - intersecting or touching
308 if ((v[0] * v[0] + v[1] * v[1] + v[2] * v[2]) < eps * eps)
309 break;
310 }
311
312 if (k == maxk)
313 throw std::runtime_error("GJK error - max iteration limit reached");
314
315 return v;
316}
317
318} // namespace dolfinx::geometry
Geometry data structures and algorithms.
Definition BoundingBoxTree.h:21
std::array< T, 3 > compute_distance_gjk(std::span< const T > p, std::span< const T > q)
Compute the distance between two convex bodies p and q, each defined by a set of points.
Definition gjk.h:256
auto norm(const V &x, Norm type=Norm::l2)
Definition Vector.h:268