| 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 testConcurrentIncrementalSmoother.cpp |
| 14 | * @brief Unit tests for the Concurrent Batch Smoother |
| 15 | * @author Stephen Williams (swilliams8@gatech.edu) |
| 16 | * @date Jan 5, 2013 |
| 17 | */ |
| 18 | |
| 19 | #include <gtsam_unstable/nonlinear/ConcurrentIncrementalSmoother.h> |
| 20 | #include <gtsam/nonlinear/PriorFactor.h> |
| 21 | #include <gtsam/slam/BetweenFactor.h> |
| 22 | #include <gtsam/nonlinear/ISAM2.h> |
| 23 | #include <gtsam/nonlinear/LevenbergMarquardtOptimizer.h> |
| 24 | #include <gtsam/nonlinear/NonlinearFactorGraph.h> |
| 25 | #include <gtsam/nonlinear/LinearContainerFactor.h> |
| 26 | #include <gtsam/nonlinear/Values.h> |
| 27 | #include <gtsam/inference/Symbol.h> |
| 28 | #include <gtsam/inference/Key.h> |
| 29 | #include <gtsam/inference/Ordering.h> |
| 30 | #include <gtsam/inference/JunctionTree.h> |
| 31 | #include <gtsam/geometry/Pose3.h> |
| 32 | #include <gtsam/base/TestableAssertions.h> |
| 33 | #include <CppUnitLite/TestHarness.h> |
| 34 | |
| 35 | using namespace std; |
| 36 | using namespace gtsam; |
| 37 | |
| 38 | namespace { |
| 39 | |
| 40 | // Set up initial pose, odometry difference, loop closure difference, and initialization errors |
| 41 | const Pose3 poseInitial; |
| 42 | const Pose3 poseOdometry( Rot3::RzRyRx(xyz: Vector3(0.05, 0.10, -0.75)), Point3(1.0, -0.25, 0.10) ); |
| 43 | const Pose3 poseError( Rot3::RzRyRx(xyz: Vector3(0.01, 0.02, -0.1)), Point3(0.05, -0.05, 0.02) ); |
| 44 | |
| 45 | // Set up noise models for the factors |
| 46 | const SharedDiagonal noisePrior = noiseModel::Isotropic::Sigma(dim: 6, sigma: 0.10); |
| 47 | const SharedDiagonal noiseOdometery = noiseModel::Diagonal::Sigmas(sigmas: (Vector(6) << 0.1, 0.1, 0.1, 0.5, 0.5, 0.5).finished()); |
| 48 | const SharedDiagonal noiseLoop = noiseModel::Diagonal::Sigmas(sigmas: (Vector(6) << 0.25, 0.25, 0.25, 1.0, 1.0, 1.0).finished()); |
| 49 | |
| 50 | /* ************************************************************************* */ |
| 51 | Values BatchOptimize(const NonlinearFactorGraph& graph, const Values& theta, int maxIter = 100) { |
| 52 | |
| 53 | // Create an L-M optimizer |
| 54 | ISAM2Params parameters; |
| 55 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 56 | // parameters.maxIterations = maxIter; |
| 57 | // parameters.lambdaUpperBound = 1; |
| 58 | // parameters.lambdaInitial = 1; |
| 59 | // parameters.verbosity = NonlinearOptimizerParams::ERROR; |
| 60 | // parameters.verbosityLM = ISAM2Params::DAMPED; |
| 61 | // parameters.linearSolverType = NonlinearOptimizerParams::MULTIFRONTAL_QR; |
| 62 | |
| 63 | ISAM2 optimizer(parameters); |
| 64 | optimizer.update( newFactors: graph, newTheta: theta ); |
| 65 | Values result = optimizer.calculateEstimate(); |
| 66 | return result; |
| 67 | |
| 68 | } |
| 69 | |
| 70 | } // end namespace |
| 71 | |
| 72 | |
| 73 | |
| 74 | |
| 75 | /* ************************************************************************* */ |
| 76 | TEST( ConcurrentIncrementalSmootherDL, equals ) |
| 77 | { |
| 78 | // TODO: Test 'equals' more vigorously |
| 79 | |
| 80 | // Create a Concurrent Batch Smoother |
| 81 | ISAM2Params parameters1; |
| 82 | parameters1.optimizationParams = ISAM2DoglegParams(); |
| 83 | ConcurrentIncrementalSmoother smoother1(parameters1); |
| 84 | |
| 85 | // Create an identical Concurrent Batch Smoother |
| 86 | ISAM2Params parameters2; |
| 87 | parameters2.optimizationParams = ISAM2DoglegParams(); |
| 88 | ConcurrentIncrementalSmoother smoother2(parameters2); |
| 89 | |
| 90 | // Create a different Concurrent Batch Smoother |
| 91 | ISAM2Params parameters3; |
| 92 | parameters3.optimizationParams = ISAM2DoglegParams(); |
| 93 | // parameters3.maxIterations = 1; |
| 94 | ConcurrentIncrementalSmoother smoother3(parameters3); |
| 95 | |
| 96 | CHECK(assert_equal(smoother1, smoother1)); |
| 97 | CHECK(assert_equal(smoother1, smoother2)); |
| 98 | // CHECK(assert_inequal(smoother1, smoother3)); |
| 99 | } |
| 100 | |
| 101 | /* ************************************************************************* */ |
| 102 | TEST( ConcurrentIncrementalSmootherDL, getFactors ) |
| 103 | { |
| 104 | // Create a Concurrent Batch Smoother |
| 105 | ISAM2Params parameters; |
| 106 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 107 | ConcurrentIncrementalSmoother smoother(parameters); |
| 108 | |
| 109 | // Expected graph is empty |
| 110 | NonlinearFactorGraph expected1; |
| 111 | // Get actual graph |
| 112 | NonlinearFactorGraph actual1 = smoother.getFactors(); |
| 113 | // Check |
| 114 | CHECK(assert_equal(expected1, actual1)); |
| 115 | |
| 116 | // Add some factors to the smoother |
| 117 | NonlinearFactorGraph newFactors1; |
| 118 | newFactors1.addPrior(key: 1, prior: poseInitial, model: noisePrior); |
| 119 | newFactors1.push_back(factor: BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery)); |
| 120 | Values newValues1; |
| 121 | newValues1.insert(j: 1, val: Pose3()); |
| 122 | newValues1.insert(j: 2, val: newValues1.at<Pose3>(j: 1).compose(g: poseOdometry)); |
| 123 | smoother.update(newFactors: newFactors1, newTheta: newValues1); |
| 124 | |
| 125 | // Expected graph |
| 126 | NonlinearFactorGraph expected2; |
| 127 | expected2.push_back(container: newFactors1); |
| 128 | // Get actual graph |
| 129 | NonlinearFactorGraph actual2 = smoother.getFactors(); |
| 130 | // Check |
| 131 | CHECK(assert_equal(expected2, actual2)); |
| 132 | |
| 133 | // Add some more factors to the smoother |
| 134 | NonlinearFactorGraph newFactors2; |
| 135 | newFactors2.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 136 | newFactors2.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 137 | Values newValues2; |
| 138 | newValues2.insert(j: 3, val: newValues1.at<Pose3>(j: 2).compose(g: poseOdometry)); |
| 139 | newValues2.insert(j: 4, val: newValues2.at<Pose3>(j: 3).compose(g: poseOdometry)); |
| 140 | smoother.update(newFactors: newFactors2, newTheta: newValues2); |
| 141 | |
| 142 | // Expected graph |
| 143 | NonlinearFactorGraph expected3; |
| 144 | expected3.push_back(container: newFactors1); |
| 145 | expected3.push_back(container: newFactors2); |
| 146 | // Get actual graph |
| 147 | NonlinearFactorGraph actual3 = smoother.getFactors(); |
| 148 | // Check |
| 149 | CHECK(assert_equal(expected3, actual3)); |
| 150 | } |
| 151 | |
| 152 | /* ************************************************************************* */ |
| 153 | TEST( ConcurrentIncrementalSmootherDL, getLinearizationPoint ) |
| 154 | { |
| 155 | // Create a Concurrent Batch Smoother |
| 156 | ISAM2Params parameters; |
| 157 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 158 | ConcurrentIncrementalSmoother smoother(parameters); |
| 159 | |
| 160 | // Expected values is empty |
| 161 | Values expected1; |
| 162 | // Get Linearization Point |
| 163 | Values actual1 = smoother.getLinearizationPoint(); |
| 164 | // Check |
| 165 | CHECK(assert_equal(expected1, actual1)); |
| 166 | |
| 167 | // Add some factors to the smoother |
| 168 | NonlinearFactorGraph newFactors1; |
| 169 | newFactors1.addPrior(key: 1, prior: poseInitial, model: noisePrior); |
| 170 | newFactors1.push_back(factor: BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery)); |
| 171 | Values newValues1; |
| 172 | newValues1.insert(j: 1, val: Pose3()); |
| 173 | newValues1.insert(j: 2, val: newValues1.at<Pose3>(j: 1).compose(g: poseOdometry)); |
| 174 | smoother.update(newFactors: newFactors1, newTheta: newValues1); |
| 175 | |
| 176 | // Expected values is equivalent to the provided values only because the provided linearization points were optimal |
| 177 | Values expected2; |
| 178 | expected2.insert(values: newValues1); |
| 179 | // Get actual linearization point |
| 180 | Values actual2 = smoother.getLinearizationPoint(); |
| 181 | // Check |
| 182 | CHECK(assert_equal(expected2, actual2)); |
| 183 | |
| 184 | // Add some more factors to the smoother |
| 185 | NonlinearFactorGraph newFactors2; |
| 186 | newFactors2.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 187 | newFactors2.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 188 | Values newValues2; |
| 189 | newValues2.insert(j: 3, val: newValues1.at<Pose3>(j: 2).compose(g: poseOdometry)); |
| 190 | newValues2.insert(j: 4, val: newValues2.at<Pose3>(j: 3).compose(g: poseOdometry)); |
| 191 | smoother.update(newFactors: newFactors2, newTheta: newValues2); |
| 192 | |
| 193 | // Expected values is equivalent to the provided values only because the provided linearization points were optimal |
| 194 | Values expected3; |
| 195 | expected3.insert(values: newValues1); |
| 196 | expected3.insert(values: newValues2); |
| 197 | // Get actual linearization point |
| 198 | Values actual3 = smoother.getLinearizationPoint(); |
| 199 | // Check |
| 200 | CHECK(assert_equal(expected3, actual3)); |
| 201 | } |
| 202 | |
| 203 | /* ************************************************************************* */ |
| 204 | TEST( ConcurrentIncrementalSmootherDL, getDelta ) |
| 205 | { |
| 206 | // TODO: Think about how to check ordering... |
| 207 | } |
| 208 | |
| 209 | /* ************************************************************************* */ |
| 210 | TEST( ConcurrentIncrementalSmootherDL, calculateEstimate ) |
| 211 | { |
| 212 | // Create a Concurrent Batch Smoother |
| 213 | ISAM2Params parameters; |
| 214 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 215 | ConcurrentIncrementalSmoother smoother(parameters); |
| 216 | |
| 217 | // Expected values is empty |
| 218 | Values expected1; |
| 219 | // Get Linearization Point |
| 220 | Values actual1 = smoother.calculateEstimate(); |
| 221 | // Check |
| 222 | CHECK(assert_equal(expected1, actual1)); |
| 223 | |
| 224 | // Add some factors to the smoother |
| 225 | NonlinearFactorGraph newFactors2; |
| 226 | newFactors2.addPrior(key: 1, prior: poseInitial, model: noisePrior); |
| 227 | newFactors2.push_back(factor: BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery)); |
| 228 | Values newValues2; |
| 229 | newValues2.insert(j: 1, val: Pose3().compose(g: poseError)); |
| 230 | newValues2.insert(j: 2, val: newValues2.at<Pose3>(j: 1).compose(g: poseOdometry).compose(g: poseError)); |
| 231 | smoother.update(newFactors: newFactors2, newTheta: newValues2); |
| 232 | |
| 233 | // Expected values from full batch optimization |
| 234 | NonlinearFactorGraph allFactors2; |
| 235 | allFactors2.push_back(container: newFactors2); |
| 236 | Values allValues2; |
| 237 | allValues2.insert(values: newValues2); |
| 238 | Values expected2 = BatchOptimize(graph: allFactors2, theta: allValues2); |
| 239 | // Get actual linearization point |
| 240 | Values actual2 = smoother.calculateEstimate(); |
| 241 | // Check |
| 242 | CHECK(assert_equal(expected2, actual2, 1e-6)); |
| 243 | |
| 244 | // Add some more factors to the smoother |
| 245 | NonlinearFactorGraph newFactors3; |
| 246 | newFactors3.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 247 | newFactors3.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 248 | Values newValues3; |
| 249 | newValues3.insert(j: 3, val: newValues2.at<Pose3>(j: 2).compose(g: poseOdometry).compose(g: poseError)); |
| 250 | newValues3.insert(j: 4, val: newValues3.at<Pose3>(j: 3).compose(g: poseOdometry).compose(g: poseError)); |
| 251 | smoother.update(newFactors: newFactors3, newTheta: newValues3); |
| 252 | |
| 253 | // Expected values from full batch optimization |
| 254 | NonlinearFactorGraph allFactors3; |
| 255 | allFactors3.push_back(container: newFactors2); |
| 256 | allFactors3.push_back(container: newFactors3); |
| 257 | Values allValues3; |
| 258 | allValues3.insert(values: newValues2); |
| 259 | allValues3.insert(values: newValues3); |
| 260 | Values expected3 = BatchOptimize(graph: allFactors3, theta: allValues3); |
| 261 | // Get actual linearization point |
| 262 | Values actual3 = smoother.calculateEstimate(); |
| 263 | // Check |
| 264 | CHECK(assert_equal(expected3, actual3, 1e-6)); |
| 265 | |
| 266 | // Also check the single-variable version |
| 267 | Pose3 expectedPose1 = expected3.at<Pose3>(j: 1); |
| 268 | Pose3 expectedPose2 = expected3.at<Pose3>(j: 2); |
| 269 | Pose3 expectedPose3 = expected3.at<Pose3>(j: 3); |
| 270 | Pose3 expectedPose4 = expected3.at<Pose3>(j: 4); |
| 271 | // Also check the single-variable version |
| 272 | Pose3 actualPose1 = smoother.calculateEstimate<Pose3>(key: 1); |
| 273 | Pose3 actualPose2 = smoother.calculateEstimate<Pose3>(key: 2); |
| 274 | Pose3 actualPose3 = smoother.calculateEstimate<Pose3>(key: 3); |
| 275 | Pose3 actualPose4 = smoother.calculateEstimate<Pose3>(key: 4); |
| 276 | // Check |
| 277 | CHECK(assert_equal(expectedPose1, actualPose1, 1e-6)); |
| 278 | CHECK(assert_equal(expectedPose2, actualPose2, 1e-6)); |
| 279 | CHECK(assert_equal(expectedPose3, actualPose3, 1e-6)); |
| 280 | CHECK(assert_equal(expectedPose4, actualPose4, 1e-6)); |
| 281 | } |
| 282 | |
| 283 | /* ************************************************************************* */ |
| 284 | TEST( ConcurrentIncrementalSmootherDL, update_empty ) |
| 285 | { |
| 286 | // Create a set of optimizer parameters |
| 287 | ISAM2Params parameters; |
| 288 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 289 | // Create a Concurrent Batch Smoother |
| 290 | ConcurrentIncrementalSmoother smoother(parameters); |
| 291 | |
| 292 | // Call update |
| 293 | smoother.update(); |
| 294 | } |
| 295 | |
| 296 | /* ************************************************************************* */ |
| 297 | TEST( ConcurrentIncrementalSmootherDL, update_multiple ) |
| 298 | { |
| 299 | // Create a Concurrent Batch Smoother |
| 300 | ISAM2Params parameters; |
| 301 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 302 | ConcurrentIncrementalSmoother smoother(parameters); |
| 303 | |
| 304 | // Expected values is empty |
| 305 | Values expected1; |
| 306 | // Get Linearization Point |
| 307 | Values actual1 = smoother.calculateEstimate(); |
| 308 | // Check |
| 309 | CHECK(assert_equal(expected1, actual1)); |
| 310 | |
| 311 | // Add some factors to the smoother |
| 312 | NonlinearFactorGraph newFactors2; |
| 313 | newFactors2.addPrior(key: 1, prior: poseInitial, model: noisePrior); |
| 314 | newFactors2.push_back(factor: BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery)); |
| 315 | Values newValues2; |
| 316 | newValues2.insert(j: 1, val: Pose3().compose(g: poseError)); |
| 317 | newValues2.insert(j: 2, val: newValues2.at<Pose3>(j: 1).compose(g: poseOdometry).compose(g: poseError)); |
| 318 | smoother.update(newFactors: newFactors2, newTheta: newValues2); |
| 319 | |
| 320 | // Expected values from full batch optimization |
| 321 | NonlinearFactorGraph allFactors2; |
| 322 | allFactors2.push_back(container: newFactors2); |
| 323 | Values allValues2; |
| 324 | allValues2.insert(values: newValues2); |
| 325 | Values expected2 = BatchOptimize(graph: allFactors2, theta: allValues2); |
| 326 | // Get actual linearization point |
| 327 | Values actual2 = smoother.calculateEstimate(); |
| 328 | // Check |
| 329 | CHECK(assert_equal(expected2, actual2, 1e-6)); |
| 330 | |
| 331 | // Add some more factors to the smoother |
| 332 | NonlinearFactorGraph newFactors3; |
| 333 | newFactors3.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 334 | newFactors3.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 335 | Values newValues3; |
| 336 | newValues3.insert(j: 3, val: newValues2.at<Pose3>(j: 2).compose(g: poseOdometry).compose(g: poseError)); |
| 337 | newValues3.insert(j: 4, val: newValues3.at<Pose3>(j: 3).compose(g: poseOdometry).compose(g: poseError)); |
| 338 | smoother.update(newFactors: newFactors3, newTheta: newValues3); |
| 339 | |
| 340 | // Expected values from full batch optimization |
| 341 | NonlinearFactorGraph allFactors3; |
| 342 | allFactors3.push_back(container: newFactors2); |
| 343 | allFactors3.push_back(container: newFactors3); |
| 344 | Values allValues3; |
| 345 | allValues3.insert(values: newValues2); |
| 346 | allValues3.insert(values: newValues3); |
| 347 | Values expected3 = BatchOptimize(graph: allFactors3, theta: allValues3); |
| 348 | // Get actual linearization point |
| 349 | Values actual3 = smoother.calculateEstimate(); |
| 350 | // Check |
| 351 | CHECK(assert_equal(expected3, actual3, 1e-6)); |
| 352 | } |
| 353 | |
| 354 | /* ************************************************************************* */ |
| 355 | TEST( ConcurrentIncrementalSmootherDL, synchronize_empty ) |
| 356 | { |
| 357 | // Create a set of optimizer parameters |
| 358 | ISAM2Params parameters; |
| 359 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 360 | // Create a Concurrent Batch Smoother |
| 361 | ConcurrentIncrementalSmoother smoother(parameters); |
| 362 | |
| 363 | // Create empty containers *from* the filter |
| 364 | NonlinearFactorGraph smootherFactors, filterSumarization; |
| 365 | Values smootherValues, filterSeparatorValues; |
| 366 | |
| 367 | // Create expected values: these will be empty for this case |
| 368 | NonlinearFactorGraph expectedSmootherSummarization; |
| 369 | Values expectedSmootherSeparatorValues; |
| 370 | |
| 371 | // Synchronize |
| 372 | NonlinearFactorGraph actualSmootherSummarization; |
| 373 | Values actualSmootherSeparatorValues; |
| 374 | smoother.presync(); |
| 375 | smoother.getSummarizedFactors(summarizedFactors&: actualSmootherSummarization, separatorValues&: actualSmootherSeparatorValues); |
| 376 | smoother.synchronize(smootherFactors, smootherValues, summarizedFactors: filterSumarization, separatorValues: filterSeparatorValues); |
| 377 | smoother.postsync(); |
| 378 | |
| 379 | // Check |
| 380 | CHECK(assert_equal(expectedSmootherSummarization, actualSmootherSummarization, 1e-6)); |
| 381 | CHECK(assert_equal(expectedSmootherSeparatorValues, actualSmootherSeparatorValues, 1e-6)); |
| 382 | } |
| 383 | |
| 384 | /* ************************************************************************* */ |
| 385 | TEST( ConcurrentIncrementalSmootherDL, synchronize_1 ) |
| 386 | { |
| 387 | // Create a set of optimizer parameters |
| 388 | ISAM2Params parameters; |
| 389 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 390 | // parameters.maxIterations = 1; |
| 391 | |
| 392 | // Create a Concurrent Batch Smoother |
| 393 | ConcurrentIncrementalSmoother smoother(parameters); |
| 394 | |
| 395 | // Create a simple separator *from* the filter |
| 396 | NonlinearFactorGraph smootherFactors, filterSumarization; |
| 397 | Values smootherValues, filterSeparatorValues; |
| 398 | filterSeparatorValues.insert(j: 1, val: Pose3().compose(g: poseError)); |
| 399 | filterSeparatorValues.insert(j: 2, val: filterSeparatorValues.at<Pose3>(j: 1).compose(g: poseOdometry).compose(g: poseError)); |
| 400 | |
| 401 | filterSumarization.push_back(factor: LinearContainerFactor(PriorFactor<Pose3>(1, poseInitial, noisePrior).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 402 | filterSumarization.push_back(factor: LinearContainerFactor(BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 403 | |
| 404 | // Create expected values: the smoother output will be empty for this case |
| 405 | NonlinearFactorGraph expectedSmootherSummarization; |
| 406 | Values expectedSmootherSeparatorValues; |
| 407 | |
| 408 | NonlinearFactorGraph actualSmootherSummarization; |
| 409 | Values actualSmootherSeparatorValues; |
| 410 | smoother.presync(); |
| 411 | smoother.getSummarizedFactors(summarizedFactors&: actualSmootherSummarization, separatorValues&: actualSmootherSeparatorValues); |
| 412 | smoother.synchronize(smootherFactors, smootherValues, summarizedFactors: filterSumarization, separatorValues: filterSeparatorValues); |
| 413 | smoother.postsync(); |
| 414 | |
| 415 | // Check |
| 416 | CHECK(assert_equal(expectedSmootherSummarization, actualSmootherSummarization, 1e-6)); |
| 417 | CHECK(assert_equal(expectedSmootherSeparatorValues, actualSmootherSeparatorValues, 1e-6)); |
| 418 | |
| 419 | |
| 420 | // Update the smoother |
| 421 | smoother.update(); |
| 422 | |
| 423 | // Check the factor graph. It should contain only the filter-provided factors |
| 424 | NonlinearFactorGraph expectedFactorGraph = filterSumarization; |
| 425 | NonlinearFactorGraph actualFactorGraph = smoother.getFactors(); |
| 426 | CHECK(assert_equal(expectedFactorGraph, actualFactorGraph, 1e-6)); |
| 427 | |
| 428 | // Check the optimized value of smoother state |
| 429 | NonlinearFactorGraph allFactors; |
| 430 | allFactors.push_back(container: filterSumarization); |
| 431 | Values allValues; |
| 432 | allValues.insert(values: filterSeparatorValues); |
| 433 | Values expectedValues = BatchOptimize(graph: allFactors, theta: allValues,maxIter: 1); |
| 434 | Values actualValues = smoother.calculateEstimate(); |
| 435 | CHECK(assert_equal(expectedValues, actualValues, 1e-6)); |
| 436 | |
| 437 | // Check the linearization point. The separator should remain identical to the filter provided values |
| 438 | Values expectedLinearizationPoint = filterSeparatorValues; |
| 439 | Values actualLinearizationPoint = smoother.getLinearizationPoint(); |
| 440 | CHECK(assert_equal(expectedLinearizationPoint, actualLinearizationPoint, 1e-6)); |
| 441 | } |
| 442 | |
| 443 | |
| 444 | /* ************************************************************************* */ |
| 445 | TEST( ConcurrentIncrementalSmootherDL, synchronize_2 ) |
| 446 | { |
| 447 | // Create a set of optimizer parameters |
| 448 | ISAM2Params parameters; |
| 449 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 450 | // parameters.maxIterations = 1; |
| 451 | // parameters.lambdaUpperBound = 1; |
| 452 | // parameters.lambdaInitial = 1; |
| 453 | // parameters.verbosity = NonlinearOptimizerParams::ERROR; |
| 454 | // parameters.verbosityLM = ISAM2Params::DAMPED; |
| 455 | |
| 456 | // Create a Concurrent Batch Smoother |
| 457 | ConcurrentIncrementalSmoother smoother(parameters); |
| 458 | |
| 459 | // Create a separator and cached smoother factors *from* the filter |
| 460 | NonlinearFactorGraph smootherFactors, filterSumarization; |
| 461 | Values smootherValues, filterSeparatorValues; |
| 462 | |
| 463 | filterSeparatorValues.insert(j: 1, val: Pose3().compose(g: poseError)); |
| 464 | filterSeparatorValues.insert(j: 2, val: filterSeparatorValues.at<Pose3>(j: 1).compose(g: poseOdometry).compose(g: poseError)); |
| 465 | filterSumarization.push_back(factor: LinearContainerFactor(PriorFactor<Pose3>(1, poseInitial, noisePrior).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 466 | filterSumarization.push_back(factor: LinearContainerFactor(BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 467 | smootherFactors.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 468 | smootherFactors.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 469 | smootherValues.insert(j: 3, val: filterSeparatorValues.at<Pose3>(j: 2).compose(g: poseOdometry).compose(g: poseError)); |
| 470 | smootherValues.insert(j: 4, val: smootherValues.at<Pose3>(j: 3).compose(g: poseOdometry).compose(g: poseError)); |
| 471 | |
| 472 | // Create expected values: the smoother output will be empty for this case |
| 473 | NonlinearFactorGraph expectedSmootherSummarization; |
| 474 | Values expectedSmootherSeparatorValues; |
| 475 | |
| 476 | NonlinearFactorGraph actualSmootherSummarization; |
| 477 | Values actualSmootherSeparatorValues; |
| 478 | smoother.presync(); |
| 479 | smoother.getSummarizedFactors(summarizedFactors&: actualSmootherSummarization, separatorValues&: actualSmootherSeparatorValues); |
| 480 | smoother.synchronize(smootherFactors, smootherValues, summarizedFactors: filterSumarization, separatorValues: filterSeparatorValues); |
| 481 | smoother.postsync(); |
| 482 | |
| 483 | // Check |
| 484 | CHECK(assert_equal(expectedSmootherSummarization, actualSmootherSummarization, 1e-6)); |
| 485 | CHECK(assert_equal(expectedSmootherSeparatorValues, actualSmootherSeparatorValues, 1e-6)); |
| 486 | |
| 487 | |
| 488 | // Update the smoother |
| 489 | smoother.update(); |
| 490 | |
| 491 | // Check the factor graph. It should contain only the filter-provided factors |
| 492 | NonlinearFactorGraph expectedFactorGraph; |
| 493 | expectedFactorGraph.push_back(container: smootherFactors); |
| 494 | expectedFactorGraph.push_back(container: filterSumarization); |
| 495 | NonlinearFactorGraph actualFactorGraph = smoother.getFactors(); |
| 496 | CHECK(assert_equal(expectedFactorGraph, actualFactorGraph, 1e-6)); |
| 497 | |
| 498 | // Check the optimized value of smoother state |
| 499 | NonlinearFactorGraph allFactors; |
| 500 | allFactors.push_back(container: filterSumarization); |
| 501 | allFactors.push_back(container: smootherFactors); |
| 502 | Values allValues; |
| 503 | allValues.insert(values: filterSeparatorValues); |
| 504 | allValues.insert(values: smootherValues); |
| 505 | // allValues.print("Batch LinPoint:\n"); |
| 506 | Values expectedValues = BatchOptimize(graph: allFactors, theta: allValues, maxIter: 1); |
| 507 | Values actualValues = smoother.calculateEstimate(); |
| 508 | CHECK(assert_equal(expectedValues, actualValues, 1e-6)); |
| 509 | |
| 510 | // Check the linearization point. The separator should remain identical to the filter provided values, but the others should be the optimal values |
| 511 | // Isam2 is changing internally the linearization points, so the following check is done only on the separator variables |
| 512 | // Values expectedLinearizationPoint = BatchOptimize(allFactors, allValues, 1); |
| 513 | Values expectedLinearizationPoint = filterSeparatorValues; |
| 514 | Values actualLinearizationPoint; |
| 515 | for(const auto key: filterSeparatorValues.keys()) { |
| 516 | actualLinearizationPoint.insert(j: key, val: smoother.getLinearizationPoint().at(j: key)); |
| 517 | } |
| 518 | CHECK(assert_equal(expectedLinearizationPoint, actualLinearizationPoint, 1e-6)); |
| 519 | } |
| 520 | |
| 521 | |
| 522 | |
| 523 | /* ************************************************************************* */ |
| 524 | TEST( ConcurrentIncrementalSmootherDL, synchronize_3 ) |
| 525 | { |
| 526 | // Create a set of optimizer parameters |
| 527 | ISAM2Params parameters; |
| 528 | parameters.optimizationParams = ISAM2DoglegParams(); |
| 529 | // parameters.maxIterations = 1; |
| 530 | // parameters.lambdaUpperBound = 1; |
| 531 | // parameters.lambdaInitial = 1; |
| 532 | // parameters.verbosity = NonlinearOptimizerParams::ERROR; |
| 533 | // parameters.verbosityLM = ISAM2Params::DAMPED; |
| 534 | |
| 535 | // Create a Concurrent Batch Smoother |
| 536 | ConcurrentIncrementalSmoother smoother(parameters); |
| 537 | |
| 538 | // Create a separator and cached smoother factors *from* the filter |
| 539 | NonlinearFactorGraph smootherFactors, filterSumarization; |
| 540 | Values smootherValues, filterSeparatorValues; |
| 541 | |
| 542 | filterSeparatorValues.insert(j: 1, val: Pose3().compose(g: poseError)); |
| 543 | filterSeparatorValues.insert(j: 2, val: filterSeparatorValues.at<Pose3>(j: 1).compose(g: poseOdometry).compose(g: poseError)); |
| 544 | filterSumarization.push_back(factor: LinearContainerFactor(PriorFactor<Pose3>(1, poseInitial, noisePrior).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 545 | filterSumarization.push_back(factor: LinearContainerFactor(BetweenFactor<Pose3>(1, 2, poseOdometry, noiseOdometery).linearize(x: filterSeparatorValues), filterSeparatorValues)); |
| 546 | smootherFactors.push_back(factor: BetweenFactor<Pose3>(2, 3, poseOdometry, noiseOdometery)); |
| 547 | smootherFactors.push_back(factor: BetweenFactor<Pose3>(3, 4, poseOdometry, noiseOdometery)); |
| 548 | smootherFactors.addPrior(key: 4, prior: poseInitial, model: noisePrior); |
| 549 | smootherValues.insert(j: 3, val: filterSeparatorValues.at<Pose3>(j: 2).compose(g: poseOdometry).compose(g: poseError)); |
| 550 | smootherValues.insert(j: 4, val: smootherValues.at<Pose3>(j: 3).compose(g: poseOdometry).compose(g: poseError)); |
| 551 | |
| 552 | // Create expected values: the smoother output will be empty for this case |
| 553 | NonlinearFactorGraph expectedSmootherSummarization; |
| 554 | Values expectedSmootherSeparatorValues; |
| 555 | |
| 556 | NonlinearFactorGraph actualSmootherSummarization; |
| 557 | Values actualSmootherSeparatorValues; |
| 558 | smoother.presync(); |
| 559 | smoother.getSummarizedFactors(summarizedFactors&: actualSmootherSummarization, separatorValues&: actualSmootherSeparatorValues); |
| 560 | smoother.synchronize(smootherFactors, smootherValues, summarizedFactors: filterSumarization, separatorValues: filterSeparatorValues); |
| 561 | smoother.postsync(); |
| 562 | |
| 563 | // Check |
| 564 | CHECK(assert_equal(expectedSmootherSummarization, actualSmootherSummarization, 1e-6)); |
| 565 | CHECK(assert_equal(expectedSmootherSeparatorValues, actualSmootherSeparatorValues, 1e-6)); |
| 566 | |
| 567 | |
| 568 | // Update the smoother |
| 569 | smoother.update(); |
| 570 | |
| 571 | smoother.presync(); |
| 572 | smoother.getSummarizedFactors(summarizedFactors&: actualSmootherSummarization, separatorValues&: actualSmootherSeparatorValues); |
| 573 | |
| 574 | // Check the optimized value of smoother state |
| 575 | NonlinearFactorGraph allFactors = smootherFactors; |
| 576 | Values allValues = smoother.getLinearizationPoint(); |
| 577 | |
| 578 | GaussianFactorGraph::shared_ptr LinFactorGraph = allFactors.linearize(linearizationPoint: allValues); |
| 579 | // GaussianSequentialSolver GSS = GaussianSequentialSolver(*LinFactorGraph); |
| 580 | // GaussianBayesNet::shared_ptr GBNsptr = GSS.eliminate(); |
| 581 | |
| 582 | KeySet allkeys = LinFactorGraph->keys(); |
| 583 | for(const auto key: filterSeparatorValues.keys()) { |
| 584 | allkeys.erase(x: key); |
| 585 | } |
| 586 | KeyVector variables(allkeys.begin(), allkeys.end()); |
| 587 | const auto [bn, fg] = LinFactorGraph->eliminatePartialSequential(variables, function: EliminateCholesky); |
| 588 | |
| 589 | expectedSmootherSummarization.resize(size: 0); |
| 590 | for(const GaussianFactor::shared_ptr& factor: *fg) { |
| 591 | expectedSmootherSummarization.push_back(factor: LinearContainerFactor(factor, allValues)); |
| 592 | } |
| 593 | |
| 594 | CHECK(assert_equal(expectedSmootherSummarization, actualSmootherSummarization, 1e-6)); |
| 595 | |
| 596 | } |
| 597 | |
| 598 | /* ************************************************************************* */ |
| 599 | int main() { TestResult tr; return TestRegistry::runAllTests(result&: tr);} |
| 600 | /* ************************************************************************* */ |
| 601 | |