| 1 | /* ---------------------------------------------------------------------------- |
| 2 | |
| 3 | * GTSAM Copyright 2010, Georgia Tech Research Corporation, |
| 4 | * Atlanta, Georgia 30332-0415 |
| 5 | * All Rights Reserved |
| 6 | * Authors: Frank Dellaert, et al. (see THANKS for the full author list) |
| 7 | |
| 8 | * See LICENSE for the license information |
| 9 | |
| 10 | * -------------------------------------------------------------------------- */ |
| 11 | |
| 12 | /** |
| 13 | * @file smallExample.h |
| 14 | * @brief Create small example with two poses and one landmark |
| 15 | * @brief smallExample |
| 16 | * @author Carlos Nieto |
| 17 | */ |
| 18 | |
| 19 | // \callgraph |
| 20 | |
| 21 | |
| 22 | #pragma once |
| 23 | |
| 24 | #include <tests/simulated2D.h> |
| 25 | #include <gtsam/inference/Symbol.h> |
| 26 | #include <gtsam/nonlinear/NonlinearFactorGraph.h> |
| 27 | #include <gtsam/linear/GaussianBayesNet.h> |
| 28 | #include <gtsam/linear/GaussianFactorGraph.h> |
| 29 | |
| 30 | namespace gtsam { |
| 31 | namespace example { |
| 32 | |
| 33 | /** |
| 34 | * Create small example for non-linear factor graph |
| 35 | */ |
| 36 | // inline std::shared_ptr<const NonlinearFactorGraph> sharedNonlinearFactorGraph(); |
| 37 | // inline NonlinearFactorGraph createNonlinearFactorGraph(); |
| 38 | |
| 39 | /** |
| 40 | * Create values structure to go with it |
| 41 | * The ground truth values structure for the example above |
| 42 | */ |
| 43 | // inline Values createValues(); |
| 44 | |
| 45 | /** Vector Values equivalent */ |
| 46 | // inline VectorValues createVectorValues(); |
| 47 | |
| 48 | /** |
| 49 | * create a noisy values structure for a nonlinear factor graph |
| 50 | */ |
| 51 | // inline std::shared_ptr<const Values> sharedNoisyValues(); |
| 52 | // inline Values createNoisyValues(); |
| 53 | |
| 54 | /** |
| 55 | * Zero delta config |
| 56 | */ |
| 57 | // inline VectorValues createZeroDelta(); |
| 58 | |
| 59 | /** |
| 60 | * Delta config that, when added to noisyValues, returns the ground truth |
| 61 | */ |
| 62 | // inline VectorValues createCorrectDelta(); |
| 63 | |
| 64 | /** |
| 65 | * create a linear factor graph |
| 66 | * The non-linear graph above evaluated at NoisyValues |
| 67 | */ |
| 68 | // inline GaussianFactorGraph createGaussianFactorGraph(); |
| 69 | |
| 70 | /** |
| 71 | * create small Chordal Bayes Net x <- y |
| 72 | */ |
| 73 | // inline GaussianBayesNet createSmallGaussianBayesNet(); |
| 74 | |
| 75 | /** |
| 76 | * Create really non-linear factor graph (cos/sin) |
| 77 | */ |
| 78 | // inline std::shared_ptr<const NonlinearFactorGraph> |
| 79 | //sharedReallyNonlinearFactorGraph(); |
| 80 | // inline NonlinearFactorGraph createReallyNonlinearFactorGraph(); |
| 81 | |
| 82 | /** |
| 83 | * Create a full nonlinear smoother |
| 84 | * @param T number of time-steps |
| 85 | */ |
| 86 | // inline std::pair<NonlinearFactorGraph, Values> createNonlinearSmoother(int T); |
| 87 | |
| 88 | /** |
| 89 | * Create a Kalman smoother by linearizing a non-linear factor graph |
| 90 | * @param T number of time-steps |
| 91 | */ |
| 92 | // inline GaussianFactorGraph createSmoother(int T); |
| 93 | |
| 94 | /* ******************************************************* */ |
| 95 | // Linear Constrained Examples |
| 96 | /* ******************************************************* */ |
| 97 | |
| 98 | /** |
| 99 | * Creates a simple constrained graph with one linear factor and |
| 100 | * one binary equality constraint that sets x = y |
| 101 | */ |
| 102 | // inline GaussianFactorGraph createSimpleConstraintGraph(); |
| 103 | // inline VectorValues createSimpleConstraintValues(); |
| 104 | |
| 105 | /** |
| 106 | * Creates a simple constrained graph with one linear factor and |
| 107 | * one binary constraint. |
| 108 | */ |
| 109 | // inline GaussianFactorGraph createSingleConstraintGraph(); |
| 110 | // inline VectorValues createSingleConstraintValues(); |
| 111 | |
| 112 | /** |
| 113 | * Creates a constrained graph with a linear factor and two |
| 114 | * binary constraints that share a node |
| 115 | */ |
| 116 | // inline GaussianFactorGraph createMultiConstraintGraph(); |
| 117 | // inline VectorValues createMultiConstraintValues(); |
| 118 | |
| 119 | /* ******************************************************* */ |
| 120 | // Planar graph with easy subtree for SubgraphPreconditioner |
| 121 | /* ******************************************************* */ |
| 122 | |
| 123 | /* |
| 124 | * Create factor graph with N^2 nodes, for example for N=3 |
| 125 | * x13-x23-x33 |
| 126 | * | | | |
| 127 | * x12-x22-x32 |
| 128 | * | | | |
| 129 | * -x11-x21-x31 |
| 130 | * with x11 clamped at (1,1), and others related by 2D odometry. |
| 131 | */ |
| 132 | // inline std::pair<GaussianFactorGraph, VectorValues> planarGraph(size_t N); |
| 133 | |
| 134 | /* |
| 135 | * Create canonical ordering for planar graph that also works for tree |
| 136 | * With x11 the root, e.g. for N=3 |
| 137 | * x33 x23 x13 x32 x22 x12 x31 x21 x11 |
| 138 | */ |
| 139 | // inline Ordering planarOrdering(size_t N); |
| 140 | |
| 141 | /* |
| 142 | * Split graph into tree and loop closing constraints, e.g., with N=3 |
| 143 | * x13-x23-x33 |
| 144 | * | |
| 145 | * x12-x22-x32 |
| 146 | * | |
| 147 | * -x11-x21-x31 |
| 148 | */ |
| 149 | // inline std::pair<GaussianFactorGraph, GaussianFactorGraph > splitOffPlanarTree( |
| 150 | // size_t N, const GaussianFactorGraph& original); |
| 151 | |
| 152 | |
| 153 | |
| 154 | // Implementations |
| 155 | |
| 156 | // using namespace gtsam::noiseModel; |
| 157 | |
| 158 | namespace impl { |
| 159 | typedef std::shared_ptr<NonlinearFactor> shared_nlf; |
| 160 | |
| 161 | static SharedDiagonal kSigma1_0 = noiseModel::Isotropic::Sigma(dim: 2,sigma: 1.0); |
| 162 | static SharedDiagonal kSigma0_1 = noiseModel::Isotropic::Sigma(dim: 2,sigma: 0.1); |
| 163 | static SharedDiagonal kSigma0_2 = noiseModel::Isotropic::Sigma(dim: 2,sigma: 0.2); |
| 164 | static SharedDiagonal kConstrainedModel = noiseModel::Constrained::All(dim: 2); |
| 165 | |
| 166 | static const Key _l1_=0, _x1_=1, _x2_=2; |
| 167 | static const Key _x_=0, _y_=1, _z_=2; |
| 168 | } // \namespace impl |
| 169 | |
| 170 | |
| 171 | /* ************************************************************************* */ |
| 172 | inline std::shared_ptr<const NonlinearFactorGraph> |
| 173 | sharedNonlinearFactorGraph(const SharedNoiseModel &noiseModel1 = impl::kSigma0_1, |
| 174 | const SharedNoiseModel &noiseModel2 = impl::kSigma0_2) { |
| 175 | using namespace impl; |
| 176 | using symbol_shorthand::L; |
| 177 | using symbol_shorthand::X; |
| 178 | // Create |
| 179 | std::shared_ptr<NonlinearFactorGraph> nlfg(new NonlinearFactorGraph); |
| 180 | |
| 181 | // prior on x1 |
| 182 | Point2 mu(0, 0); |
| 183 | shared_nlf f1(new simulated2D::Prior(mu, noiseModel1, X(j: 1))); |
| 184 | nlfg->push_back(factor: f1); |
| 185 | |
| 186 | // odometry between x1 and x2 |
| 187 | Point2 z2(1.5, 0); |
| 188 | shared_nlf f2(new simulated2D::Odometry(z2, noiseModel1, X(j: 1), X(j: 2))); |
| 189 | nlfg->push_back(factor: f2); |
| 190 | |
| 191 | // measurement between x1 and l1 |
| 192 | Point2 z3(0, -1); |
| 193 | shared_nlf f3(new simulated2D::Measurement(z3, noiseModel2, X(j: 1), L(j: 1))); |
| 194 | nlfg->push_back(factor: f3); |
| 195 | |
| 196 | // measurement between x2 and l1 |
| 197 | Point2 z4(-1.5, -1.); |
| 198 | shared_nlf f4(new simulated2D::Measurement(z4, noiseModel2, X(j: 2), L(j: 1))); |
| 199 | nlfg->push_back(factor: f4); |
| 200 | |
| 201 | return nlfg; |
| 202 | } |
| 203 | |
| 204 | /* ************************************************************************* */ |
| 205 | inline NonlinearFactorGraph |
| 206 | createNonlinearFactorGraph(const SharedNoiseModel &noiseModel1 = impl::kSigma0_1, |
| 207 | const SharedNoiseModel &noiseModel2 = impl::kSigma0_2) { |
| 208 | return *sharedNonlinearFactorGraph(noiseModel1, noiseModel2); |
| 209 | } |
| 210 | |
| 211 | /* ************************************************************************* */ |
| 212 | inline Values createValues() { |
| 213 | using symbol_shorthand::X; |
| 214 | using symbol_shorthand::L; |
| 215 | Values c; |
| 216 | c.insert(j: X(j: 1), val: Point2(0.0, 0.0)); |
| 217 | c.insert(j: X(j: 2), val: Point2(1.5, 0.0)); |
| 218 | c.insert(j: L(j: 1), val: Point2(0.0, -1.0)); |
| 219 | return c; |
| 220 | } |
| 221 | |
| 222 | /* ************************************************************************* */ |
| 223 | inline VectorValues createVectorValues() { |
| 224 | using namespace impl; |
| 225 | VectorValues c {{_l1_, Vector2(0.0, -1.0)}, |
| 226 | {_x1_, Vector2(0.0, 0.0)}, |
| 227 | {_x2_, Vector2(1.5, 0.0)}}; |
| 228 | return c; |
| 229 | } |
| 230 | |
| 231 | /* ************************************************************************* */ |
| 232 | inline std::shared_ptr<const Values> sharedNoisyValues() { |
| 233 | using symbol_shorthand::X; |
| 234 | using symbol_shorthand::L; |
| 235 | std::shared_ptr<Values> c(new Values); |
| 236 | c->insert(j: X(j: 1), val: Point2(0.1, 0.1)); |
| 237 | c->insert(j: X(j: 2), val: Point2(1.4, 0.2)); |
| 238 | c->insert(j: L(j: 1), val: Point2(0.1, -1.1)); |
| 239 | return c; |
| 240 | } |
| 241 | |
| 242 | /* ************************************************************************* */ |
| 243 | inline Values createNoisyValues() { |
| 244 | return *sharedNoisyValues(); |
| 245 | } |
| 246 | |
| 247 | /* ************************************************************************* */ |
| 248 | inline VectorValues createCorrectDelta() { |
| 249 | using symbol_shorthand::X; |
| 250 | using symbol_shorthand::L; |
| 251 | VectorValues c; |
| 252 | c.insert(j: L(j: 1), value: Vector2(-0.1, 0.1)); |
| 253 | c.insert(j: X(j: 1), value: Vector2(-0.1, -0.1)); |
| 254 | c.insert(j: X(j: 2), value: Vector2(0.1, -0.2)); |
| 255 | return c; |
| 256 | } |
| 257 | |
| 258 | /* ************************************************************************* */ |
| 259 | inline VectorValues createZeroDelta() { |
| 260 | using symbol_shorthand::X; |
| 261 | using symbol_shorthand::L; |
| 262 | VectorValues c; |
| 263 | c.insert(j: L(j: 1), value: Z_2x1); |
| 264 | c.insert(j: X(j: 1), value: Z_2x1); |
| 265 | c.insert(j: X(j: 2), value: Z_2x1); |
| 266 | return c; |
| 267 | } |
| 268 | |
| 269 | /* ************************************************************************* */ |
| 270 | inline GaussianFactorGraph createGaussianFactorGraph() { |
| 271 | using symbol_shorthand::X; |
| 272 | using symbol_shorthand::L; |
| 273 | // Create empty graph |
| 274 | GaussianFactorGraph fg; |
| 275 | |
| 276 | // linearized prior on x1: c[_x1_]+x1=0 i.e. x1=-c[_x1_] |
| 277 | fg.emplace_shared<JacobianFactor>(args: X(j: 1), args: 10*I_2x2, args: -1.0*Vector::Ones(newSize: 2)); |
| 278 | |
| 279 | // odometry between x1 and x2: x2-x1=[0.2;-0.1] |
| 280 | fg.emplace_shared<JacobianFactor>(args: X(j: 1), args: -10*I_2x2, args: X(j: 2), args: 10*I_2x2, args: Vector2(2.0, -1.0)); |
| 281 | |
| 282 | // measurement between x1 and l1: l1-x1=[0.0;0.2] |
| 283 | fg.emplace_shared<JacobianFactor>(args: X(j: 1), args: -5*I_2x2, args: L(j: 1), args: 5*I_2x2, args: Vector2(0.0, 1.0)); |
| 284 | |
| 285 | // measurement between x2 and l1: l1-x2=[-0.2;0.3] |
| 286 | fg.emplace_shared<JacobianFactor>(args: X(j: 2), args: -5*I_2x2, args: L(j: 1), args: 5*I_2x2, args: Vector2(-1.0, 1.5)); |
| 287 | |
| 288 | return fg; |
| 289 | } |
| 290 | |
| 291 | /* ************************************************************************* */ |
| 292 | /** create small Chordal Bayes Net x <- y |
| 293 | * x y d |
| 294 | * 1 1 9 |
| 295 | * 1 5 |
| 296 | */ |
| 297 | inline GaussianBayesNet createSmallGaussianBayesNet() { |
| 298 | using namespace impl; |
| 299 | Matrix R11 = (Matrix(1, 1) << 1.0).finished(), S12 = (Matrix(1, 1) << 1.0).finished(); |
| 300 | Matrix R22 = (Matrix(1, 1) << 1.0).finished(); |
| 301 | Vector d1(1), d2(1); |
| 302 | d1(0) = 9; |
| 303 | d2(0) = 5; |
| 304 | |
| 305 | // define nodes and specify in reverse topological sort (i.e. parents last) |
| 306 | GaussianConditional::shared_ptr Px_y(new GaussianConditional(_x_, d1, R11, _y_, S12)); |
| 307 | GaussianConditional::shared_ptr Py(new GaussianConditional(_y_, d2, R22)); |
| 308 | GaussianBayesNet cbn; |
| 309 | cbn.push_back(factor: Px_y); |
| 310 | cbn.push_back(factor: Py); |
| 311 | |
| 312 | return cbn; |
| 313 | } |
| 314 | |
| 315 | /* ************************************************************************* */ |
| 316 | // Some nonlinear functions to optimize |
| 317 | /* ************************************************************************* */ |
| 318 | namespace smallOptimize { |
| 319 | |
| 320 | inline Point2 h(const Point2& v) { |
| 321 | return Point2(cos(x: v.x()), sin(x: v.y())); |
| 322 | } |
| 323 | |
| 324 | inline Matrix H(const Point2& v) { |
| 325 | return (Matrix(2, 2) << |
| 326 | -sin(x: v.x()), 0.0, |
| 327 | 0.0, cos(x: v.y())).finished(); |
| 328 | } |
| 329 | |
| 330 | struct UnaryFactor: public gtsam::NoiseModelFactorN<Point2> { |
| 331 | |
| 332 | // Provide access to the Matrix& version of evaluateError: |
| 333 | using gtsam::NoiseModelFactor1<Point2>::evaluateError; |
| 334 | |
| 335 | Point2 z_; |
| 336 | |
| 337 | UnaryFactor(const Point2& z, const SharedNoiseModel& model, Key key) : |
| 338 | gtsam::NoiseModelFactorN<Point2>(model, key), z_(z) { |
| 339 | } |
| 340 | |
| 341 | Vector evaluateError(const Point2& x, OptionalMatrixType A) const override { |
| 342 | if (A) *A = H(v: x); |
| 343 | return (h(v: x) - z_); |
| 344 | } |
| 345 | |
| 346 | gtsam::NonlinearFactor::shared_ptr clone() const override { |
| 347 | return std::static_pointer_cast<gtsam::NonlinearFactor>( |
| 348 | r: gtsam::NonlinearFactor::shared_ptr(new UnaryFactor(*this))); } |
| 349 | }; |
| 350 | |
| 351 | } |
| 352 | |
| 353 | /* ************************************************************************* */ |
| 354 | inline NonlinearFactorGraph nonlinearFactorGraphWithGivenSigma(const double sigma) { |
| 355 | using symbol_shorthand::X; |
| 356 | using symbol_shorthand::L; |
| 357 | std::shared_ptr<NonlinearFactorGraph> fg(new NonlinearFactorGraph); |
| 358 | Point2 z(1.0, 0.0); |
| 359 | std::shared_ptr<smallOptimize::UnaryFactor> factor( |
| 360 | new smallOptimize::UnaryFactor(z, noiseModel::Isotropic::Sigma(dim: 2,sigma), X(j: 1))); |
| 361 | fg->push_back(factor); |
| 362 | return *fg; |
| 363 | } |
| 364 | |
| 365 | /* ************************************************************************* */ |
| 366 | inline std::shared_ptr<const NonlinearFactorGraph> sharedReallyNonlinearFactorGraph() { |
| 367 | using symbol_shorthand::X; |
| 368 | using symbol_shorthand::L; |
| 369 | std::shared_ptr<NonlinearFactorGraph> fg(new NonlinearFactorGraph); |
| 370 | Point2 z(1.0, 0.0); |
| 371 | double sigma = 0.1; |
| 372 | std::shared_ptr<smallOptimize::UnaryFactor> factor( |
| 373 | new smallOptimize::UnaryFactor(z, noiseModel::Isotropic::Sigma(dim: 2,sigma), X(j: 1))); |
| 374 | fg->push_back(factor); |
| 375 | return fg; |
| 376 | } |
| 377 | |
| 378 | inline NonlinearFactorGraph createReallyNonlinearFactorGraph() { |
| 379 | return *sharedReallyNonlinearFactorGraph(); |
| 380 | } |
| 381 | |
| 382 | /* ************************************************************************* */ |
| 383 | inline NonlinearFactorGraph sharedNonRobustFactorGraphWithOutliers() { |
| 384 | using symbol_shorthand::X; |
| 385 | std::shared_ptr<NonlinearFactorGraph> fg(new NonlinearFactorGraph); |
| 386 | Point2 z(0.0, 0.0); |
| 387 | double sigma = 0.1; |
| 388 | |
| 389 | std::shared_ptr<PriorFactor<Point2>> factor( |
| 390 | new PriorFactor<Point2>(X(j: 1), z, noiseModel::Isotropic::Sigma(dim: 2,sigma))); |
| 391 | // 3 noiseless inliers |
| 392 | fg->push_back(factor); |
| 393 | fg->push_back(factor); |
| 394 | fg->push_back(factor); |
| 395 | |
| 396 | // 1 outlier |
| 397 | Point2 z_out(1.0, 0.0); |
| 398 | std::shared_ptr<PriorFactor<Point2>> factor_out( |
| 399 | new PriorFactor<Point2>(X(j: 1), z_out, noiseModel::Isotropic::Sigma(dim: 2,sigma))); |
| 400 | fg->push_back(factor: factor_out); |
| 401 | |
| 402 | return *fg; |
| 403 | } |
| 404 | |
| 405 | /* ************************************************************************* */ |
| 406 | inline NonlinearFactorGraph sharedRobustFactorGraphWithOutliers() { |
| 407 | using symbol_shorthand::X; |
| 408 | std::shared_ptr<NonlinearFactorGraph> fg(new NonlinearFactorGraph); |
| 409 | Point2 z(0.0, 0.0); |
| 410 | double sigma = 0.1; |
| 411 | auto gmNoise = noiseModel::Robust::Create( |
| 412 | robust: noiseModel::mEstimator::GemanMcClure::Create(k: 1.0), noise: noiseModel::Isotropic::Sigma(dim: 2,sigma)); |
| 413 | std::shared_ptr<PriorFactor<Point2>> factor( |
| 414 | new PriorFactor<Point2>(X(j: 1), z, gmNoise)); |
| 415 | // 3 noiseless inliers |
| 416 | fg->push_back(factor); |
| 417 | fg->push_back(factor); |
| 418 | fg->push_back(factor); |
| 419 | |
| 420 | // 1 outlier |
| 421 | Point2 z_out(1.0, 0.0); |
| 422 | std::shared_ptr<PriorFactor<Point2>> factor_out( |
| 423 | new PriorFactor<Point2>(X(j: 1), z_out, gmNoise)); |
| 424 | fg->push_back(factor: factor_out); |
| 425 | |
| 426 | return *fg; |
| 427 | } |
| 428 | |
| 429 | |
| 430 | /* ************************************************************************* */ |
| 431 | inline std::pair<NonlinearFactorGraph, Values> createNonlinearSmoother(int T) { |
| 432 | using namespace impl; |
| 433 | using symbol_shorthand::X; |
| 434 | using symbol_shorthand::L; |
| 435 | |
| 436 | // Create |
| 437 | NonlinearFactorGraph nlfg; |
| 438 | Values poses; |
| 439 | |
| 440 | // prior on x1 |
| 441 | Point2 x1(1.0, 0.0); |
| 442 | shared_nlf prior(new simulated2D::Prior(x1, kSigma1_0, X(j: 1))); |
| 443 | nlfg.push_back(factor: prior); |
| 444 | poses.insert(j: X(j: 1), val: x1); |
| 445 | |
| 446 | for (int t = 2; t <= T; t++) { |
| 447 | // odometry between x_t and x_{t-1} |
| 448 | Point2 odo(1.0, 0.0); |
| 449 | shared_nlf odometry(new simulated2D::Odometry(odo, kSigma1_0, X(j: t - 1), X(j: t))); |
| 450 | nlfg.push_back(factor: odometry); |
| 451 | |
| 452 | // measurement on x_t is like perfect GPS |
| 453 | Point2 xt(t, 0); |
| 454 | shared_nlf measurement(new simulated2D::Prior(xt, kSigma1_0, X(j: t))); |
| 455 | nlfg.push_back(factor: measurement); |
| 456 | |
| 457 | // initial estimate |
| 458 | poses.insert(j: X(j: t), val: xt); |
| 459 | } |
| 460 | |
| 461 | return std::make_pair(x&: nlfg, y&: poses); |
| 462 | } |
| 463 | |
| 464 | /* ************************************************************************* */ |
| 465 | inline GaussianFactorGraph createSmoother(int T) { |
| 466 | const auto [nlfg, poses] = createNonlinearSmoother(T); |
| 467 | return *nlfg.linearize(linearizationPoint: poses); |
| 468 | } |
| 469 | |
| 470 | /* ************************************************************************* */ |
| 471 | inline GaussianFactorGraph createSimpleConstraintGraph() { |
| 472 | using namespace impl; |
| 473 | // create unary factor |
| 474 | // prior on _x_, mean = [1,-1], sigma=0.1 |
| 475 | Matrix Ax = I_2x2; |
| 476 | Vector b1(2); |
| 477 | b1(0) = 1.0; |
| 478 | b1(1) = -1.0; |
| 479 | JacobianFactor::shared_ptr f1(new JacobianFactor(_x_, Ax, b1, kSigma0_1)); |
| 480 | |
| 481 | // create binary constraint factor |
| 482 | // between _x_ and _y_, that is going to be the only factor on _y_ |
| 483 | // |1 0||x_1| + |-1 0||y_1| = |0| |
| 484 | // |0 1||x_2| | 0 -1||y_2| |0| |
| 485 | Matrix Ax1 = I_2x2; |
| 486 | Matrix Ay1 = I_2x2 * -1; |
| 487 | Vector b2 = Vector2(0.0, 0.0); |
| 488 | JacobianFactor::shared_ptr f2(new JacobianFactor(_x_, Ax1, _y_, Ay1, b2, |
| 489 | kConstrainedModel)); |
| 490 | |
| 491 | // construct the graph |
| 492 | GaussianFactorGraph fg; |
| 493 | fg.push_back(factor: f1); |
| 494 | fg.push_back(factor: f2); |
| 495 | |
| 496 | return fg; |
| 497 | } |
| 498 | |
| 499 | /* ************************************************************************* */ |
| 500 | inline VectorValues createSimpleConstraintValues() { |
| 501 | using namespace impl; |
| 502 | using symbol_shorthand::X; |
| 503 | using symbol_shorthand::L; |
| 504 | VectorValues config; |
| 505 | Vector v = Vector2(1.0, -1.0); |
| 506 | config.insert(j: _x_, value: v); |
| 507 | config.insert(j: _y_, value: v); |
| 508 | return config; |
| 509 | } |
| 510 | |
| 511 | /* ************************************************************************* */ |
| 512 | inline GaussianFactorGraph createSingleConstraintGraph() { |
| 513 | using namespace impl; |
| 514 | // create unary factor |
| 515 | // prior on _x_, mean = [1,-1], sigma=0.1 |
| 516 | Matrix Ax = I_2x2; |
| 517 | Vector b1(2); |
| 518 | b1(0) = 1.0; |
| 519 | b1(1) = -1.0; |
| 520 | //GaussianFactor::shared_ptr f1(new JacobianFactor(_x_, kSigma0_1->Whiten(Ax), kSigma0_1->whiten(b1), kSigma0_1)); |
| 521 | JacobianFactor::shared_ptr f1(new JacobianFactor(_x_, Ax, b1, kSigma0_1)); |
| 522 | |
| 523 | // create binary constraint factor |
| 524 | // between _x_ and _y_, that is going to be the only factor on _y_ |
| 525 | // |1 2||x_1| + |10 0||y_1| = |1| |
| 526 | // |2 1||x_2| |0 10||y_2| |2| |
| 527 | Matrix Ax1(2, 2); |
| 528 | Ax1(0, 0) = 1.0; |
| 529 | Ax1(0, 1) = 2.0; |
| 530 | Ax1(1, 0) = 2.0; |
| 531 | Ax1(1, 1) = 1.0; |
| 532 | Matrix Ay1 = I_2x2 * 10; |
| 533 | Vector b2 = Vector2(1.0, 2.0); |
| 534 | JacobianFactor::shared_ptr f2(new JacobianFactor(_x_, Ax1, _y_, Ay1, b2, |
| 535 | kConstrainedModel)); |
| 536 | |
| 537 | // construct the graph |
| 538 | GaussianFactorGraph fg; |
| 539 | fg.push_back(factor: f1); |
| 540 | fg.push_back(factor: f2); |
| 541 | |
| 542 | return fg; |
| 543 | } |
| 544 | |
| 545 | /* ************************************************************************* */ |
| 546 | inline VectorValues createSingleConstraintValues() { |
| 547 | using namespace impl; |
| 548 | VectorValues config{{_x_, Vector2(1.0, -1.0)}, {_y_, Vector2(0.2, 0.1)}}; |
| 549 | return config; |
| 550 | } |
| 551 | |
| 552 | /* ************************************************************************* */ |
| 553 | inline GaussianFactorGraph createMultiConstraintGraph() { |
| 554 | using namespace impl; |
| 555 | // unary factor 1 |
| 556 | Matrix A = I_2x2; |
| 557 | Vector b = Vector2(-2.0, 2.0); |
| 558 | JacobianFactor::shared_ptr lf1(new JacobianFactor(_x_, A, b, kSigma0_1)); |
| 559 | |
| 560 | // constraint 1 |
| 561 | Matrix A11(2, 2); |
| 562 | A11(0, 0) = 1.0; |
| 563 | A11(0, 1) = 2.0; |
| 564 | A11(1, 0) = 2.0; |
| 565 | A11(1, 1) = 1.0; |
| 566 | |
| 567 | Matrix A12(2, 2); |
| 568 | A12(0, 0) = 10.0; |
| 569 | A12(0, 1) = 0.0; |
| 570 | A12(1, 0) = 0.0; |
| 571 | A12(1, 1) = 10.0; |
| 572 | |
| 573 | Vector b1(2); |
| 574 | b1(0) = 1.0; |
| 575 | b1(1) = 2.0; |
| 576 | JacobianFactor::shared_ptr lc1(new JacobianFactor(_x_, A11, _y_, A12, b1, |
| 577 | kConstrainedModel)); |
| 578 | |
| 579 | // constraint 2 |
| 580 | Matrix A21(2, 2); |
| 581 | A21(0, 0) = 3.0; |
| 582 | A21(0, 1) = 4.0; |
| 583 | A21(1, 0) = -1.0; |
| 584 | A21(1, 1) = -2.0; |
| 585 | |
| 586 | Matrix A22(2, 2); |
| 587 | A22(0, 0) = 1.0; |
| 588 | A22(0, 1) = 1.0; |
| 589 | A22(1, 0) = 1.0; |
| 590 | A22(1, 1) = 2.0; |
| 591 | |
| 592 | Vector b2(2); |
| 593 | b2(0) = 3.0; |
| 594 | b2(1) = 4.0; |
| 595 | JacobianFactor::shared_ptr lc2(new JacobianFactor(_x_, A21, _z_, A22, b2, |
| 596 | kConstrainedModel)); |
| 597 | |
| 598 | // construct the graph |
| 599 | GaussianFactorGraph fg; |
| 600 | fg.push_back(factor: lf1); |
| 601 | fg.push_back(factor: lc1); |
| 602 | fg.push_back(factor: lc2); |
| 603 | |
| 604 | return fg; |
| 605 | } |
| 606 | |
| 607 | /* ************************************************************************* */ |
| 608 | inline VectorValues createMultiConstraintValues() { |
| 609 | using namespace impl; |
| 610 | VectorValues config{{_x_, Vector2(-2.0, 2.0)}, |
| 611 | {_y_, Vector2(-0.1, 0.4)}, |
| 612 | {_z_, Vector2(-4.0, 5.0)}}; |
| 613 | return config; |
| 614 | } |
| 615 | |
| 616 | /* ************************************************************************* */ |
| 617 | // Create key for simulated planar graph |
| 618 | namespace impl { |
| 619 | inline Symbol key(size_t x, size_t y) { |
| 620 | using symbol_shorthand::X; |
| 621 | return X(j: 1000*x+y); |
| 622 | } |
| 623 | } // \namespace impl |
| 624 | |
| 625 | /* ************************************************************************* */ |
| 626 | inline std::pair<GaussianFactorGraph, VectorValues> planarGraph(size_t N) { |
| 627 | using namespace impl; |
| 628 | |
| 629 | // create empty graph |
| 630 | NonlinearFactorGraph nlfg; |
| 631 | |
| 632 | // Create almost hard constraint on x11, sigma=0 will work for PCG not for normal |
| 633 | shared_nlf constraint(new simulated2D::Prior(Point2(1.0, 1.0), noiseModel::Isotropic::Sigma(dim: 2,sigma: 1e-3), key(x: 1,y: 1))); |
| 634 | nlfg.push_back(factor: constraint); |
| 635 | |
| 636 | // Create horizontal constraints, 1...N*(N-1) |
| 637 | Point2 z1(1.0, 0.0); // move right |
| 638 | for (size_t x = 1; x < N; x++) |
| 639 | for (size_t y = 1; y <= N; y++) { |
| 640 | shared_nlf f(new simulated2D::Odometry(z1, noiseModel::Isotropic::Sigma(dim: 2,sigma: 0.01), key(x, y), key(x: x + 1, y))); |
| 641 | nlfg.push_back(factor: f); |
| 642 | } |
| 643 | |
| 644 | // Create vertical constraints, N*(N-1)+1..2*N*(N-1) |
| 645 | Point2 z2(0.0, 1.0); // move up |
| 646 | for (size_t x = 1; x <= N; x++) |
| 647 | for (size_t y = 1; y < N; y++) { |
| 648 | shared_nlf f(new simulated2D::Odometry(z2, noiseModel::Isotropic::Sigma(dim: 2,sigma: 0.01), key(x, y), key(x, y: y + 1))); |
| 649 | nlfg.push_back(factor: f); |
| 650 | } |
| 651 | |
| 652 | // Create linearization and ground xtrue config |
| 653 | Values zeros; |
| 654 | for (size_t x = 1; x <= N; x++) |
| 655 | for (size_t y = 1; y <= N; y++) |
| 656 | zeros.insert(j: key(x, y), val: Point2(0,0)); |
| 657 | VectorValues xtrue; |
| 658 | for (size_t x = 1; x <= N; x++) |
| 659 | for (size_t y = 1; y <= N; y++) |
| 660 | xtrue.insert(j: key(x, y), value: Point2((double)x, (double)y)); |
| 661 | |
| 662 | // linearize around zero |
| 663 | std::shared_ptr<GaussianFactorGraph> gfg = nlfg.linearize(linearizationPoint: zeros); |
| 664 | return std::make_pair(x&: *gfg, y&: xtrue); |
| 665 | } |
| 666 | |
| 667 | /* ************************************************************************* */ |
| 668 | inline Ordering planarOrdering(size_t N) { |
| 669 | Ordering ordering; |
| 670 | for (size_t y = N; y >= 1; y--) |
| 671 | for (size_t x = N; x >= 1; x--) |
| 672 | ordering.push_back(x: impl::key(x, y)); |
| 673 | return ordering; |
| 674 | } |
| 675 | |
| 676 | /* ************************************************************************* */ |
| 677 | inline std::pair<GaussianFactorGraph, GaussianFactorGraph> splitOffPlanarTree( |
| 678 | size_t N, const GaussianFactorGraph& original) { |
| 679 | GaussianFactorGraph T, C; |
| 680 | |
| 681 | // Add the x11 constraint to the tree |
| 682 | T.push_back(factor: original[0]); |
| 683 | |
| 684 | // Add all horizontal constraints to the tree |
| 685 | size_t i = 1; |
| 686 | for (size_t x = 1; x < N; x++) |
| 687 | for (size_t y = 1; y <= N; y++, i++) T.push_back(factor: original[i]); |
| 688 | |
| 689 | // Add first vertical column of constraints to T, others to C |
| 690 | for (size_t x = 1; x <= N; x++) |
| 691 | for (size_t y = 1; y < N; y++, i++) |
| 692 | if (x == 1) |
| 693 | T.push_back(factor: original[i]); |
| 694 | else |
| 695 | C.push_back(factor: original[i]); |
| 696 | |
| 697 | return std::make_pair(x&: T, y&: C); |
| 698 | } |
| 699 | |
| 700 | /* ************************************************************************* */ |
| 701 | |
| 702 | } // \namespace example |
| 703 | } // \namespace gtsam |
| 704 | |