381 lines
14 KiB
Matlab
381 lines
14 KiB
Matlab
import gtsam.*;
|
|
|
|
% Test GTSAM covariances on a graph with betweenFactors
|
|
% Optionally, you can also enable IMU factors and Camera factors
|
|
% Authors: Luca Carlone, David Jensen
|
|
% Date: 2014/4/6
|
|
|
|
clc
|
|
clear all
|
|
close all
|
|
|
|
%% Configuration
|
|
useRealData = 0; % controls whether or not to use the Real data (is available) as the ground truth traj
|
|
|
|
includeIMUFactors = 1; % if true, IMU type 1 Factors will be generated for the random trajectory
|
|
includeCameraFactors = 0; % not fully implemented yet
|
|
|
|
trajectoryLength = 4; % length of the ground truth trajectory
|
|
|
|
numMonteCarloRuns = 0;
|
|
|
|
%% Camera metadata
|
|
numberOfLandmarks = 10; % Total number of visual landmarks, used for camera factors
|
|
K = Cal3_S2(500,500,0,640/2,480/2); % Camera calibration
|
|
cameraMeasurementNoiseSigma = 1.0;
|
|
cameraMeasurementNoise = noiseModel.Isotropic.Sigma(2,cameraMeasurementNoiseSigma);
|
|
|
|
% Create landmarks
|
|
if includeCameraFactors == 1
|
|
for i = 1:numberOfLandmarks
|
|
gtLandmarkPoints(i) = Point3( ...
|
|
[rand()*20*(trajectoryLength*1.2) + 15*20; ... % uniformly distributed in the x axis along 120% of the trajectory length, starting after 15 poses
|
|
randn()*20; ... % normally distributed in the y axis with a sigma of 20
|
|
randn()*20]); % normally distributed in the z axis with a sigma of 20
|
|
end
|
|
end
|
|
|
|
%% Imu metadata
|
|
epsBias = 1e-20; % was 1e-7
|
|
zeroBias = imuBias.ConstantBias(zeros(3,1), zeros(3,1));
|
|
IMU_metadata.AccelerometerSigma = 1e-5;
|
|
IMU_metadata.GyroscopeSigma = 1e-7;
|
|
IMU_metadata.IntegrationSigma = 1e-10;
|
|
IMU_metadata.BiasAccelerometerSigma = epsBias;
|
|
IMU_metadata.BiasGyroscopeSigma = epsBias;
|
|
IMU_metadata.BiasAccOmegaInit = epsBias;
|
|
noiseVel = noiseModel.Isotropic.Sigma(3, 1e-10); % was 0.1
|
|
noiseBias = noiseModel.Isotropic.Sigma(6, epsBias);
|
|
|
|
%% Between metadata
|
|
if useRealData == 1
|
|
sigma_ang = 1e-1; sigma_cart = 1;
|
|
else
|
|
sigma_ang = 1e-2; sigma_cart = 0.1;
|
|
end
|
|
testName = sprintf('sa-%1.2g-sc-%1.2g',sigma_ang,sigma_cart)
|
|
folderName = 'results/'
|
|
|
|
noiseVectorPose = [sigma_ang; sigma_ang; sigma_ang; sigma_cart; sigma_cart; sigma_cart];
|
|
noisePose = noiseModel.Diagonal.Sigmas(noiseVectorPose);
|
|
%noisePose = noiseModel.Isotropic.Sigma(6, 1e-3);
|
|
|
|
%% Create ground truth trajectory
|
|
gtValues = Values;
|
|
gtGraph = NonlinearFactorGraph;
|
|
|
|
if useRealData == 1
|
|
subsampleStep = 20;
|
|
|
|
%% Create a ground truth trajectory from Real data (if available)
|
|
fprintf('\nUsing real data as ground truth\n');
|
|
gtScenario = load('truth_scen2.mat', 'Time', 'Lat', 'Lon', 'Alt', 'Roll', 'Pitch', 'Heading',...
|
|
'VEast', 'VNorth', 'VUp');
|
|
|
|
Org_lat = gtScenario.Lat(1);
|
|
Org_lon = gtScenario.Lon(1);
|
|
initialPositionECEF = imuSimulator.LatLonHRad_to_ECEF([gtScenario.Lat(1); gtScenario.Lon(1); gtScenario.Alt(1)]);
|
|
|
|
% Limit the trajectory length
|
|
trajectoryLength = min([length(gtScenario.Lat) trajectoryLength]);
|
|
|
|
for i=1:trajectoryLength
|
|
currentPoseKey = symbol('x', i-1);
|
|
scenarioInd = subsampleStep * (i-1) + 1
|
|
gtECEF = imuSimulator.LatLonHRad_to_ECEF([gtScenario.Lat(scenarioInd); gtScenario.Lon(scenarioInd); gtScenario.Alt(scenarioInd)]);
|
|
% truth in ENU
|
|
dX = gtECEF(1) - initialPositionECEF(1);
|
|
dY = gtECEF(2) - initialPositionECEF(2);
|
|
dZ = gtECEF(3) - initialPositionECEF(3);
|
|
[xlt, ylt, zlt] = imuSimulator.ct2ENU(dX, dY, dZ,Org_lat, Org_lon);
|
|
|
|
gtPosition = [xlt, ylt, zlt]';
|
|
gtRotation = Rot3; % Rot3.ypr(gtScenario.Heading(scenarioInd), gtScenario.Pitch(scenarioInd), gtScenario.Roll(scenarioInd));
|
|
currentPose = Pose3(gtRotation, Point3(gtPosition));
|
|
|
|
% Add values
|
|
gtValues.insert(currentPoseKey, currentPose);
|
|
|
|
if i==1 % first time step, add priors
|
|
warning('roll-pitch-yaw is different from Rodriguez')
|
|
warning('using identity rotation')
|
|
gtGraph.add(PriorFactorPose3(currentPoseKey, currentPose, noisePose));
|
|
measurements.posePrior = currentPose;
|
|
else
|
|
% Generate measurements as the current pose measured in the frame of
|
|
% the previous pose
|
|
deltaPose = prevPose.between(currentPose);
|
|
measurements.gtDeltaMatrix(i-1,:) = Pose3.Logmap(deltaPose);
|
|
|
|
% Add the factor to the factor graph
|
|
gtGraph.add(BetweenFactorPose3(currentPoseKey-1, currentPoseKey, deltaPose, noisePose));
|
|
end
|
|
prevPose = currentPose;
|
|
end
|
|
else
|
|
%% Create a random trajectory as ground truth
|
|
currentVel = [0; 0; 0]; % initial velocity (used to generate IMU measurements)
|
|
currentPose = Pose3; % initial pose % initial pose
|
|
deltaT = 0.1; % amount of time between IMU measurements
|
|
g = [0; 0; 0]; % gravity
|
|
omegaCoriolis = [0; 0; 0]; % Coriolis
|
|
|
|
unsmooth_DP = 0.5; % controls smoothness on translation norm
|
|
unsmooth_DR = 0.1; % controls smoothness on rotation norm
|
|
|
|
unsmooth_DP = 0;
|
|
unsmooth_DR = 0;
|
|
|
|
fprintf('\nCreating a random ground truth trajectory\n');
|
|
%% Add priors
|
|
currentPoseKey = symbol('x', 0);
|
|
gtValues.insert(currentPoseKey, currentPose);
|
|
% NOSIE ON PRIOR WAS TOO HIGH? Changing this fixed the indeterminant
|
|
% linear system error
|
|
gtGraph.add(PriorFactorPose3(currentPoseKey, currentPose, noiseModel.Isotropic.Sigma(6, 1e-3)));
|
|
%gtGraph.add(PriorFactorPose3(currentPoseKey, currentPose, noisePose);
|
|
|
|
if includeIMUFactors == 1
|
|
currentVelKey = symbol('v', 0);
|
|
currentBiasKey = symbol('b', 0);
|
|
gtValues.insert(currentVelKey, LieVector(currentVel));
|
|
gtValues.insert(currentBiasKey, zeroBias);
|
|
gtGraph.add(PriorFactorLieVector(currentVelKey, LieVector(currentVel), noiseVel));
|
|
gtGraph.add(PriorFactorConstantBias(currentBiasKey, zeroBias, noiseBias));
|
|
end
|
|
|
|
if includeCameraFactors == 1
|
|
pointNoiseSigma = 0.1;
|
|
pointPriorNoise = noiseModel.Isotropic.Sigma(3,pointNoiseSigma);
|
|
gtGraph.add(PriorFactorPoint3(symbol('p',1), gtLandmarkPoints(1), pointPriorNoise));
|
|
end
|
|
|
|
for i=1:trajectoryLength
|
|
currentPoseKey = symbol('x', i);
|
|
|
|
gtDeltaPosition = unsmooth_DP*randn(3,1) + [20;0;0]; % create random vector with mean = [20 0 0]
|
|
gtDeltaRotation = unsmooth_DR*randn(3,1) + [0;0;0]; % create random rotation with mean [0 0 0]
|
|
measurements.gtDeltaMatrix(i,:) = [gtDeltaRotation; gtDeltaPosition];
|
|
gtMeasurements.deltaPose = Pose3.Expmap(measurements.gtDeltaMatrix(i,:)');
|
|
|
|
% "Deduce" ground truth measurements
|
|
% deltaPose are the gt measurements - save them in some structure
|
|
currentPose = currentPose.compose(gtMeasurements.deltaPose);
|
|
gtValues.insert(currentPoseKey, currentPose);
|
|
|
|
% Add the factors to the factor graph
|
|
%gtGraph.add(BetweenFactorPose3(currentPoseKey-1, currentPoseKey, gtMeasurements.deltaPose, noisePose));
|
|
|
|
%% Add IMU factors
|
|
if includeIMUFactors == 1
|
|
currentVelKey = symbol('v', i); % not used if includeIMUFactors is false
|
|
currentBiasKey = symbol('b', i); % not used if includeIMUFactors is false
|
|
|
|
% create accel and gyro measurements based on
|
|
gtMeasurements.imu.gyro = measurements.gtDeltaMatrix(i, 1:3)'./deltaT;
|
|
% acc = (deltaPosition - initialVel * dT) * (2/dt^2)
|
|
gtMeasurements.imu.accel = (measurements.gtDeltaMatrix(i, 4:6)' - currentVel.*deltaT).*(2/(deltaT*deltaT));
|
|
% Initialize preintegration
|
|
imuMeasurement = gtsam.ImuFactorPreintegratedMeasurements(...
|
|
zeroBias, ...
|
|
IMU_metadata.AccelerometerSigma.^2 * eye(3), ...
|
|
IMU_metadata.GyroscopeSigma.^2 * eye(3), ...
|
|
IMU_metadata.IntegrationSigma.^2 * eye(3));
|
|
% Preintegrate
|
|
imuMeasurement.integrateMeasurement(gtMeasurements.imu.accel, gtMeasurements.imu.gyro, deltaT);
|
|
% Add Imu factor
|
|
gtGraph.add(ImuFactor(currentPoseKey-1, currentVelKey-1, currentPoseKey, currentVelKey, ...
|
|
currentBiasKey-1, imuMeasurement, g, omegaCoriolis));
|
|
% Add between on biases
|
|
gtGraph.add(BetweenFactorConstantBias(currentBiasKey-1, currentBiasKey, zeroBias, ...
|
|
noiseModel.Isotropic.Sigma(6, epsBias)));
|
|
% Additional prior on zerobias
|
|
gtGraph.add(PriorFactorConstantBias(currentBiasKey, zeroBias, ...
|
|
noiseModel.Isotropic.Sigma(6, epsBias)));
|
|
|
|
% update current velocity
|
|
currentVel = measurements.gtDeltaMatrix(i,4:6)'./deltaT;
|
|
gtValues.insert(currentVelKey, LieVector(currentVel));
|
|
|
|
%gtGraph.add(PriorFactorLieVector(currentVelKey, LieVector(currentVel), noiseVel));
|
|
|
|
gtValues.insert(currentBiasKey, zeroBias);
|
|
end % end of IMU factor creation
|
|
|
|
%% Add Camera factors
|
|
if includeCameraFactors == 1
|
|
% Create camera with the current pose and calibration K (specified above)
|
|
gtCamera = SimpleCamera(currentPose, K);
|
|
% Project landmarks into the camera
|
|
numSkipped = 0;
|
|
for j = 1:length(gtLandmarkPoints)
|
|
landmarkKey = symbol('p', j);
|
|
try
|
|
Z = gtCamera.project(gtLandmarkPoints(j));
|
|
%% TO-DO probably want to do some type of filtering on the measurement values, because
|
|
% they might not all be valid
|
|
gtGraph.add(GenericProjectionFactorCal3_S2(Z, cameraMeasurementNoise, currentPoseKey, landmarkKey, K));
|
|
catch
|
|
% Most likely the point is not within the camera's view, which
|
|
% is fine
|
|
numSkipped = numSkipped + 1;
|
|
end
|
|
end
|
|
%fprintf('(Pose %d) %d landmarks behind the camera\n', i, numSkipped);
|
|
end % end of Camera factor creation
|
|
end % end of trajectory length
|
|
|
|
%% Add landmark positions to the Values
|
|
if includeCameraFactors == 1
|
|
for j = 1:length(gtLandmarkPoints)
|
|
landmarkKey = symbol('p', j);
|
|
gtValues.insert(landmarkKey, gtLandmarkPoints(j));
|
|
end
|
|
end
|
|
|
|
end % end of ground truth creation
|
|
|
|
gtGraph.print(sprintf('\nGround Truth Factor graph:\n'));
|
|
gtValues.print(sprintf('\nGround Truth Values:\n '));
|
|
|
|
warning('Additional prior on zerobias')
|
|
warning('Additional PriorFactorLieVector on velocities')
|
|
|
|
% gtPoses = Values;
|
|
% for i=0:trajectoryLength
|
|
% currentPoseKey = symbol('x', i);
|
|
% currentPose = gtValues.at(currentPoseKey);
|
|
% gtPoses.insert(currentPoseKey, currentPose);
|
|
% end
|
|
|
|
figure(1)
|
|
hold on;
|
|
plot3DPoints(gtValues);
|
|
plot3DTrajectory(gtValues, '-r', [], 1, Marginals(gtGraph, gtValues));
|
|
axis equal
|
|
|
|
disp('Plotted ground truth')
|
|
|
|
for k=1:numMonteCarloRuns
|
|
fprintf('Monte Carlo Run %d.\n', k');
|
|
% create a new graph
|
|
graph = NonlinearFactorGraph;
|
|
|
|
% noisy prior
|
|
currentPoseKey = symbol('x', 0);
|
|
measurements.posePrior = currentPose;
|
|
noisyDelta = noiseVectorPose .* randn(6,1);
|
|
noisyInitialPose = Pose3.Expmap(noisyDelta);
|
|
graph.add(PriorFactorPose3(currentPoseKey, noisyInitialPose, noisePose));
|
|
|
|
for i=1:size(measurements.gtDeltaMatrix,1)
|
|
currentPoseKey = symbol('x', i);
|
|
|
|
% for each measurement: add noise and add to graph
|
|
noisyDelta = measurements.gtDeltaMatrix(i,:)' + (noiseVectorPose .* randn(6,1));
|
|
noisyDeltaPose = Pose3.Expmap(noisyDelta);
|
|
|
|
% Add the factors to the factor graph
|
|
%graph.add(BetweenFactorPose3(currentPoseKey-1, currentPoseKey, noisyDeltaPose, noisePose));
|
|
end
|
|
|
|
% optimize
|
|
optimizer = GaussNewtonOptimizer(graph, gtValues);
|
|
estimate = optimizer.optimize();
|
|
|
|
figure(1)
|
|
plot3DTrajectory(estimate, '-b');
|
|
|
|
marginals = Marginals(graph, estimate);
|
|
|
|
% for each pose in the trajectory
|
|
for i=1:size(measurements.gtDeltaMatrix,1)+1
|
|
% compute estimation errors
|
|
currentPoseKey = symbol('x', i-1);
|
|
gtPosition = gtValues.at(currentPoseKey).translation.vector;
|
|
estPosition = estimate.at(currentPoseKey).translation.vector;
|
|
estR = estimate.at(currentPoseKey).rotation.matrix;
|
|
errPosition = estPosition - gtPosition;
|
|
|
|
% compute covariances:
|
|
cov = marginals.marginalCovariance(currentPoseKey);
|
|
covPosition = estR * cov(4:6,4:6) * estR';
|
|
|
|
% compute NEES using (estimationError = estimatedValues - gtValues) and estimated covariances
|
|
NEES(k,i) = errPosition' * inv(covPosition) * errPosition; % distributed according to a Chi square with n = 3 dof
|
|
end
|
|
|
|
figure(2)
|
|
hold on
|
|
plot(NEES(k,:),'-b','LineWidth',1.5)
|
|
end
|
|
%%
|
|
ANEES = mean(NEES);
|
|
plot(ANEES,'-r','LineWidth',2)
|
|
plot(3*ones(size(ANEES,2),1),'k--'); % Expectation(ANEES) = number of dof
|
|
box on
|
|
set(gca,'Fontsize',16)
|
|
title('NEES and ANEES');
|
|
%print('-djpeg', horzcat('runs-',testName));
|
|
saveas(gcf,horzcat(folderName,'runs-',testName,'.fig'),'fig');
|
|
|
|
%%
|
|
figure(1)
|
|
box on
|
|
set(gca,'Fontsize',16)
|
|
title('Ground truth and estimates for each MC runs');
|
|
%print('-djpeg', horzcat('gt-',testName));
|
|
saveas(gcf,horzcat(folderName,'gt-',testName,'.fig'),'fig');
|
|
|
|
%% Let us compute statistics on the overall NEES
|
|
n = 3; % position vector dimension
|
|
N = numMonteCarloRuns; % number of runs
|
|
alpha = 0.01; % confidence level
|
|
|
|
% mean_value = n*N; % mean value of the Chi-square distribution
|
|
% (we divide by n * N and for this reason we expect ANEES around 1)
|
|
r1 = chi2inv(alpha, n * N) / (n * N);
|
|
r2 = chi2inv(1-alpha, n * N) / (n * N);
|
|
|
|
% output here
|
|
fprintf(1, 'r1 = %g\n', r1);
|
|
fprintf(1, 'r2 = %g\n', r2);
|
|
|
|
figure(3)
|
|
hold on
|
|
plot(ANEES/n,'-b','LineWidth',2)
|
|
plot(ones(size(ANEES,2),1),'r-');
|
|
plot(r1*ones(size(ANEES,2),1),'k-.');
|
|
plot(r2*ones(size(ANEES,2),1),'k-.');
|
|
box on
|
|
set(gca,'Fontsize',16)
|
|
title('NEES normalized by dof VS bounds');
|
|
%print('-djpeg', horzcat('ANEES-',testName));
|
|
saveas(gcf,horzcat(folderName,'ANEES-',testName,'.fig'),'fig');
|
|
|
|
logFile = horzcat(folderName,'log-',testName);
|
|
save(logFile)
|
|
|
|
%% NEES COMPUTATION (Bar-Shalom 2001, Section 5.4)
|
|
% the nees for a single experiment (i) is defined as
|
|
% NEES_i = xtilda' * inv(P) * xtilda,
|
|
% where xtilda in R^n is the estimation
|
|
% error, and P is the covariance estimated by the approach we want to test
|
|
%
|
|
% Average NEES. Given N Monte Carlo simulations, i=1,...,N, the average
|
|
% NEES is:
|
|
% ANEES = sum(NEES_i)/N
|
|
% The quantity N*ANEES is distributed according to a Chi-square
|
|
% distribution with N*n degrees of freedom.
|
|
%
|
|
% For the single run case, N=1, therefore NEES = ANEES is distributed
|
|
% according to a chi-square distribution with n degrees of freedom (e.g. n=3
|
|
% if we are testing a position estimate)
|
|
% Therefore its mean should be n (difficult to see from a single run)
|
|
% and, with probability alpha, it should hold:
|
|
%
|
|
% NEES in [r1, r2]
|
|
%
|
|
% where r1 and r2 are built from the Chi-square distribution
|
|
|