ConcurrentFilteringAndSmoothingExample.cpp

Go to the documentation of this file.

/* ----------------------------------------------------------------------------
 
 * GTSAM Copyright 2010, Georgia Tech Research Corporation,
 * Atlanta, Georgia 30332-0415
 * All Rights Reserved
 * Authors: Frank Dellaert, et al. (see THANKS for the full author list)
 
 * See LICENSE for the license information
 
 * -------------------------------------------------------------------------- */
 
// This example demonstrates the use of the Concurrent Filtering and Smoothing architecture in GTSAM unstable
#include <gtsam_unstable/nonlinear/ConcurrentBatchFilter.h>
#include <gtsam_unstable/nonlinear/ConcurrentBatchSmoother.h>
 
// We will compare the results to a similar Fixed-Lag Smoother
#include <gtsam/nonlinear/BatchFixedLagSmoother.h>
 
// In GTSAM, measurement functions are represented as 'factors'. Several common factors
// have been provided with the library for solving robotics/SLAM/Bundle Adjustment problems.
// Here we will use Between factors for the relative motion described by odometry measurements.
// Also, we will initialize the robot at the origin using a Prior factor.
#include <gtsam/slam/BetweenFactor.h>
 
// When the factors are created, we will add them to a Factor Graph. As the factors we are using
// are nonlinear factors, we will need a Nonlinear Factor Graph.
#include <gtsam/nonlinear/NonlinearFactorGraph.h>
 
// The nonlinear solvers within GTSAM are iterative solvers, meaning they linearize the
// nonlinear functions around an initial linearization point, then solve the linear system
// to update the linearization point. This happens repeatedly until the solver converges
// to a consistent set of variable values. This requires us to specify an initial guess
// for each variable, held in a Values container.
#include <gtsam/nonlinear/Values.h>
 
// We will use simple integer Keys to uniquely identify each robot pose.
#include <gtsam/inference/Key.h>
 
// We will use Pose2 variables (x, y, theta) to represent the robot positions
#include <gtsam/geometry/Pose2.h>
 
#include <iomanip>
 
using namespace std;
using namespace gtsam;
 
 
int main(int argc, char** argv) {
 
  // Define the smoother lag (in seconds)
  double lag = 2.0;
 
  // Create a Concurrent Filter and Smoother
  ConcurrentBatchFilter concurrentFilter;
  ConcurrentBatchSmoother concurrentSmoother;
 
  // And a fixed lag smoother with a short lag
  BatchFixedLagSmoother fixedlagSmoother(lag);
 
  // And a fixed lag smoother with very long lag (i.e. a full batch smoother)
  BatchFixedLagSmoother batchSmoother(1000.0);
 
 
  // Create containers to store the factors and linearization points that
  // will be sent to the smoothers
  NonlinearFactorGraph newFactors;
  Values newValues;
  FixedLagSmoother::KeyTimestampMap newTimestamps;
 
  // Create a prior on the first pose, placing it at the origin
  Pose2 priorMean(0.0, 0.0, 0.0); // prior at origin
  auto priorNoise = noiseModel::Diagonal::Sigmas(Vector3(0.3, 0.3, 0.1));
  Key priorKey = 0;
  newFactors.addPrior(priorKey, priorMean, priorNoise);
  newValues.insert(priorKey, priorMean); // Initialize the first pose at the mean of the prior
  newTimestamps[priorKey] = 0.0; // Set the timestamp associated with this key to 0.0 seconds;
 
  // Now, loop through several time steps, creating factors from different "sensors"
  // and adding them to the fixed-lag smoothers
  double deltaT = 0.25;
  for(double time = 0.0+deltaT; time <= 5.0; time += deltaT) {
 
    // Define the keys related to this timestamp
    Key previousKey(1000 * (time-deltaT));
    Key currentKey(1000 * (time));
 
    // Assign the current key to the current timestamp
    newTimestamps[currentKey] = time;
 
    // Add a guess for this pose to the new values
    // Since the robot moves forward at 2 m/s, then the position is simply: time[s]*2.0[m/s]
    // {This is not a particularly good way to guess, but this is just an example}
    Pose2 currentPose(time * 2.0, 0.0, 0.0);
    newValues.insert(currentKey, currentPose);
 
    // Add odometry factors from two different sources with different error stats
    Pose2 odometryMeasurement1 = Pose2(0.61, -0.08, 0.02);
    auto odometryNoise1 = noiseModel::Diagonal::Sigmas(Vector3(0.1, 0.1, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement1, odometryNoise1));
 
    Pose2 odometryMeasurement2 = Pose2(0.47, 0.03, 0.01);
    auto odometryNoise2 = noiseModel::Diagonal::Sigmas(Vector3(0.05, 0.05, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement2, odometryNoise2));
 
    // Unlike the fixed-lag versions, the concurrent filter implementation
    // requires the user to supply the specify which keys to move to the smoother
    FastList<Key> oldKeys;
    if(time >= lag+deltaT) {
      oldKeys.push_back(1000 * (time-lag-deltaT));
    }
 
    // Update the various inference engines
    concurrentFilter.update(newFactors, newValues, oldKeys);
    fixedlagSmoother.update(newFactors, newValues, newTimestamps);
    batchSmoother.update(newFactors, newValues, newTimestamps);
 
    // Manually synchronize the Concurrent Filter and Smoother every 1.0 s
    if(fmod(time, 1.0) < 0.01) {
      // Synchronize the Filter and Smoother
      concurrentSmoother.update();
      synchronize(concurrentFilter, concurrentSmoother);
    }
 
    // Print the optimized current pose
    cout << setprecision(5) << "Timestamp = " << time << endl;
    concurrentFilter.calculateEstimate<Pose2>(currentKey).print("Concurrent Estimate: ");
    fixedlagSmoother.calculateEstimate<Pose2>(currentKey).print("Fixed Lag Estimate:  ");
    batchSmoother.calculateEstimate<Pose2>(currentKey).print("Batch Estimate:      ");
    cout << endl;
 
    // Clear contains for the next iteration
    newTimestamps.clear();
    newValues.clear();
    newFactors.resize(0);
  }
  cout << "******************************************************************" << endl;
  cout << "All three versions should be identical." << endl;
  cout << "Adding a loop closure factor to the Batch version only." << endl;
  cout << "******************************************************************" << endl;
  cout << endl;
 
  // At the moment, all three versions produce the same output.
  // Now lets create a "loop closure" factor between the first pose and the current pose
  Key loopKey1(1000 * (0.0));
  Key loopKey2(1000 * (5.0));
  Pose2 loopMeasurement = Pose2(9.5, 1.00, 0.00);
  auto loopNoise = noiseModel::Diagonal::Sigmas(Vector3(0.5, 0.5, 0.25));
  NonlinearFactor::shared_ptr loopFactor(new BetweenFactor<Pose2>(loopKey1, loopKey2, loopMeasurement, loopNoise));
 
  // This measurement cannot be added directly to the concurrent filter because it connects a filter state to a smoother state
  // This measurement can never be added to the fixed-lag smoother, as one of the poses has been permanently marginalized out
  // This measurement can be incorporated into the full batch version though
  newFactors.push_back(loopFactor);
  batchSmoother.update(newFactors, Values(), FixedLagSmoother::KeyTimestampMap());
  newFactors.resize(0);
 
 
 
  // Continue adding odometry factors until the loop closure may be incorporated into the concurrent smoother
  for(double time = 5.0+deltaT; time <= 8.0; time += deltaT) {
 
    // Define the keys related to this timestamp
    Key previousKey(1000 * (time-deltaT));
    Key currentKey(1000 * (time));
 
    // Assign the current key to the current timestamp
    newTimestamps[currentKey] = time;
 
    // Add a guess for this pose to the new values
    // Since the robot moves forward at 2 m/s, then the position is simply: time[s]*2.0[m/s]
    // {This is not a particularly good way to guess, but this is just an example}
    Pose2 currentPose(time * 2.0, 0.0, 0.0);
    newValues.insert(currentKey, currentPose);
 
    // Add odometry factors from two different sources with different error stats
    Pose2 odometryMeasurement1 = Pose2(0.61, -0.08, 0.02);
    auto odometryNoise1 = noiseModel::Diagonal::Sigmas(Vector3(0.1, 0.1, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement1, odometryNoise1));
 
    Pose2 odometryMeasurement2 = Pose2(0.47, 0.03, 0.01);
    auto odometryNoise2 = noiseModel::Diagonal::Sigmas(Vector3(0.05, 0.05, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement2, odometryNoise2));
 
    // Unlike the fixed-lag versions, the concurrent filter implementation
    // requires the user to supply the specify which keys to marginalize
    FastList<Key> oldKeys;
    if(time >= lag+deltaT) {
      oldKeys.push_back(1000 * (time-lag-deltaT));
    }
 
    // Update the various inference engines
    concurrentFilter.update(newFactors, newValues, oldKeys);
    fixedlagSmoother.update(newFactors, newValues, newTimestamps);
    batchSmoother.update(newFactors, newValues, newTimestamps);
 
    // Manually synchronize the Concurrent Filter and Smoother every 1.0 s
    if(fmod(time, 1.0) < 0.01) {
      // Synchronize the Filter and Smoother
      concurrentSmoother.update();
      synchronize(concurrentFilter, concurrentSmoother);
    }
 
    // Print the optimized current pose
    cout << setprecision(5) << "Timestamp = " << time << endl;
    concurrentFilter.calculateEstimate<Pose2>(currentKey).print("Concurrent Estimate: ");
    fixedlagSmoother.calculateEstimate<Pose2>(currentKey).print("Fixed Lag Estimate:  ");
    batchSmoother.calculateEstimate<Pose2>(currentKey).print("Batch Estimate:      ");
    cout << endl;
 
    // Clear contains for the next iteration
    newTimestamps.clear();
    newValues.clear();
    newFactors.resize(0);
  }
  cout << "******************************************************************" << endl;
  cout << "The Concurrent system and the Fixed-Lag Smoother should be " << endl;
  cout << "the same, but the Batch version has a loop closure." << endl;
  cout << "Adding the loop closure factor to the Concurrent version." << endl;
  cout << "This will not update the Concurrent Filter until the next " << endl;
  cout << "synchronization, but the Concurrent solution should be identical " << endl;
  cout << "to the Batch solution afterwards." << endl;
  cout << "******************************************************************" << endl;
  cout << endl;
 
  // The state at 5.0s should have been transferred to the concurrent smoother at this point. Add the loop closure.
  newFactors.push_back(loopFactor);
  concurrentSmoother.update(newFactors, Values());
  newFactors.resize(0);
 
 
  // Now run for a few more seconds so the concurrent smoother and filter have to to re-sync
  // Continue adding odometry factors until the loop closure may be incorporated into the concurrent smoother
  for(double time = 8.0+deltaT; time <= 15.0; time += deltaT) {
 
    // Define the keys related to this timestamp
    Key previousKey(1000 * (time-deltaT));
    Key currentKey(1000 * (time));
 
    // Assign the current key to the current timestamp
    newTimestamps[currentKey] = time;
 
    // Add a guess for this pose to the new values
    // Since the robot moves forward at 2 m/s, then the position is simply: time[s]*2.0[m/s]
    // {This is not a particularly good way to guess, but this is just an example}
    Pose2 currentPose(time * 2.0, 0.0, 0.0);
    newValues.insert(currentKey, currentPose);
 
    // Add odometry factors from two different sources with different error stats
    Pose2 odometryMeasurement1 = Pose2(0.61, -0.08, 0.02);
    auto odometryNoise1 = noiseModel::Diagonal::Sigmas(Vector3(0.1, 0.1, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement1, odometryNoise1));
 
    Pose2 odometryMeasurement2 = Pose2(0.47, 0.03, 0.01);
    auto odometryNoise2 = noiseModel::Diagonal::Sigmas(Vector3(0.05, 0.05, 0.05));
    newFactors.push_back(BetweenFactor<Pose2>(previousKey, currentKey, odometryMeasurement2, odometryNoise2));
 
    // Unlike the fixed-lag versions, the concurrent filter implementation
    // requires the user to supply the specify which keys to marginalize
    FastList<Key> oldKeys;
    if(time >= lag+deltaT) {
      oldKeys.push_back(1000 * (time-lag-deltaT));
    }
 
    // Update the various inference engines
    concurrentFilter.update(newFactors, newValues, oldKeys);
    fixedlagSmoother.update(newFactors, newValues, newTimestamps);
    batchSmoother.update(newFactors, newValues, newTimestamps);
 
    // Manually synchronize the Concurrent Filter and Smoother every 1.0 s
    if(fmod(time, 1.0) < 0.01) {
      // Synchronize the Filter and Smoother
      concurrentSmoother.update();
      synchronize(concurrentFilter, concurrentSmoother);
      cout << "******************************************************************" << endl;
      cout << "Syncing Concurrent Filter and Smoother." << endl;
      cout << "******************************************************************" << endl;
      cout << endl;
    }
 
    // Print the optimized current pose
    cout << setprecision(5) << "Timestamp = " << time << endl;
    concurrentFilter.calculateEstimate<Pose2>(currentKey).print("Concurrent Estimate: ");
    fixedlagSmoother.calculateEstimate<Pose2>(currentKey).print("Fixed Lag Estimate:  ");
    batchSmoother.calculateEstimate<Pose2>(currentKey).print("Batch Estimate:      ");
    cout << endl;
 
    // Clear contains for the next iteration
    newTimestamps.clear();
    newValues.clear();
    newFactors.resize(0);
  }
 
 
  // And to demonstrate the fixed-lag aspect, print the keys contained in each smoother after 3.0 seconds
  cout << "After 15.0 seconds, each version contains to the following keys:" << endl;
  cout << "  Concurrent Filter Keys: " << endl;
  for(const auto key: concurrentFilter.getLinearizationPoint().keys()) {
    cout << setprecision(5) << "    Key: " << key << endl;
  }
  cout << "  Concurrent Smoother Keys: " << endl;
  for(const auto key: concurrentSmoother.getLinearizationPoint().keys()) {
    cout << setprecision(5) << "    Key: " << key << endl;
  }
  cout << "  Fixed-Lag Smoother Keys: " << endl;
  for(const auto& key_timestamp: fixedlagSmoother.timestamps()) {
    cout << setprecision(5) << "    Key: " << key_timestamp.first << endl;
  }
  cout << "  Batch Smoother Keys: " << endl;
  for(const auto& key_timestamp: batchSmoother.timestamps()) {
    cout << setprecision(5) << "    Key: " << key_timestamp.first << endl;
  }
 
  return 0;
}