initrpekf

Constant velocity range-parameterized EKF initialization

Description

filter = initrpekf(detection) configures the filter with 6 extended Kalman filters (EKFs), and the target range is assumed to be within 1e3 and 1e5 scenario units. The function configures the process noise with unit standard deviation in acceleration.

The range-parameterized extended Kalman filter (RPEKF) is a Gaussian-sum filter (trackingGSF) with multiple EKFs, each initialized at an estimated range of the target. Range-parameterization is a commonly used technique to initialize a filter from an angle-only detection.

filter = initrpekf(detection,numFilters) specifies the number of EKFs in the filter.

example

filter = initrpekf(detection,numFilters,rangeLimits) specifies the range limits of the target.

Examples

collapse all

The RPEKF is a special type of filter that can be initialized using angle-only measurements, that is, azimuth and/or elevation. When the 'Frame' is set to 'spherical' and 'HasRange' is set to 'false', the detection has [azimuth elevation] measurements. Specify the measurement parameters appropriately to define an angle-only measurement with no range information.

measParam = struct('Frame','spherical','HasRange',false,'OriginPosition',[100;10;0]);

The objectDetection class defines an interface to the angle-only detection measured by the sensor. The MeasurementParameters field of objectDetection carries information about what the sensor is measuring.

detection = objectDetection(0,[30;30],'MeasurementParameters',measParam,'MeasurementNoise',2*eye(2));

The initrpekf function uses the angle-only detection to initialize the RPEKF.

rpekf = initrpekf(detection) %#ok
rpekf = 
  trackingGSF with properties:

                 State: [6x1 double]
       StateCovariance: [6x6 double]

       TrackingFilters: {6x1 cell}
    ModelProbabilities: [6x1 double]

      MeasurementNoise: [2x2 double]

You can also initialize the RPEKF with 10 filters and to operate within the range limits of [1000, 10,000] scenario units.

rangeLimits = [1000 10000];
numFilters = 10;
rpekf = initrpekf(detection, numFilters, rangeLimits)
rpekf = 
  trackingGSF with properties:

                 State: [6x1 double]
       StateCovariance: [6x6 double]

       TrackingFilters: {10x1 cell}
    ModelProbabilities: [10x1 double]

      MeasurementNoise: [2x2 double]

You can also specify the initrpekf function as a FilterInitializationFcn to the trackerGNN object.

funcHandle = @(detection)initrpekf(detection,numFilters,rangeLimits)
funcHandle = function_handle with value:
    @(detection)initrpekf(detection,numFilters,rangeLimits)

tracker = trackerGNN('FilterInitializationFcn',funcHandle)
tracker = 
  trackerGNN with properties:

                  TrackerIndex: 0
       FilterInitializationFcn: [function_handle]
                    Assignment: 'MatchPairs'
           AssignmentThreshold: [30 Inf]
                  MaxNumTracks: 100
                 MaxNumSensors: 20

                    TrackLogic: 'History'
         ConfirmationThreshold: [2 3]
             DeletionThreshold: [5 5]

            HasCostMatrixInput: false
    HasDetectableTrackIDsInput: false
               StateParameters: [1x1 struct]

                     NumTracks: 0
            NumConfirmedTracks: 0

Visualize the filter.

tp = theaterPlot;
componentPlot = trackPlotter(tp,'DisplayName','Individual sums','MarkerFaceColor','r');
sumPlot = trackPlotter(tp,'DisplayName','Mixed State','MarkerFaceColor','g');

indFilters = rpekf.TrackingFilters;
pos = zeros(numFilters,3);
cov = zeros(3,3,numFilters);
for i = 1:numFilters
    pos(i,:) = indFilters{i}.State(1:2:end);
    cov(1:3,1:3,i) = indFilters{i}.StateCovariance(1:2:end,1:2:end);
end
componentPlot.plotTrack(pos,cov);

mixedPos = rpekf.State(1:2:end)';
mixedPosCov = rpekf.StateCovariance(1:2:end,1:2:end);
sumPlot.plotTrack(mixedPos,mixedPosCov);

Input Arguments

collapse all

Detection report, specified as an Object Detections object.

Example: detection = objectDetection(0,[1;4.5;3],'MeasurementNoise', [1.0 0 0; 0 2.0 0; 0 0 1.5])

Number of EKFs each initialized at an estimated range of the target, specified as a positive integer. When not specified, the default number of EKFs is 6.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Range limits of the target, specified as a two-element vector. The two elements in the vector represent the lower and upper limits of the target range. When not specified, the default range limits are [1e3 1e5] scenario units.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Output Arguments

collapse all

Constant velocity range-parameterized extended Kalman filter (EKF), returned as a trackingGSF object.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Introduced in R2018b