Skip to content

matlab/matlab-agentic-toolkit / skills-catalog/radar/matlab-import-tracking-data

matlab-import-tracking-data

Verified

Import raw data (CSV, XLSX, TXT, or MATLAB tables) into formats used by Sensor Fusion and Tracking Toolbox. Handles both ground truth trajectories and sensor detection data. For truth: builds trackingScenarioRecording, tuning timetable, truthlog, or converted table. For sensor data: builds task-oriented dataFormat structs (preferred) or objectDetection arrays (legacy). Use when importing flight logs, GPS logs, radar detections, IR measurements, lidar/camera bounding boxes, ADS-B data, AIS ship tracks, or any recorded data for use with trackers, filter tuning, or tracker evaluation.

references/custom-measurement-models.md

View on GitHub
# Custom Measurement Models for Legacy Trackers

Use this reference when measurements don't fit built-in models (`cvmeas`, `cameas`, etc.) and a custom measurement function + filter init pair is needed.

## When You Need Custom Models

- TDOA (time difference of arrival)
- Custom geometry not covered by standard spherical/rectangular
- Camera bounding boxes used with a legacy tracker
- Any measurement that requires domain-specific math to relate state to observation

## Custom MeasurementParameters Design

The struct can contain ANY fields. Design principles:

1. **Must carry all info the measurement function needs** to compute expected measurement from state (sensor geometry, intrinsics, baseline distances, etc.)
2. **Must include a discriminator** if multi-sensor (so the shared measurement function can branch)
3. **All detections must use the same struct fields** regardless of sensor — leave unused fields empty or with defaults

```matlab
% Example: TDOA sensor pair
mp = struct( ...
    'SensorType', 'tdoa', ...           % discriminator
    'Receiver1Position', [0;0;0], ...   % first receiver position
    'Receiver2Position', [100;0;0], ... % second receiver position
    'SpeedOfSignal', 3e8);              % propagation speed
```

## Custom Measurement Function Template

Signature for additive noise: `z = measFcn(state, mp)`

```matlab
function z = myMeasFcn(state, mp)
    % state: [x; vx; y; vy; z; vz] or your custom state vector
    % mp: your MeasurementParameters struct
    
    switch mp.SensorType
        case 'tdoa'
            pos = state([1,3,5]);  % extract position from state
            r1 = norm(pos - mp.Receiver1Position);
            r2 = norm(pos - mp.Receiver2Position);
            z = (r1 - r2) / mp.SpeedOfSignal;  % time difference
            
        case 'bearing'
            pos = state([1,3,5]) - mp.SensorPosition;
            localPos = mp.SensorRotation * pos;
            z = [atan2d(localPos(2), localPos(1));   % azimuth
                 atan2d(localPos(3), norm(localPos(1:2)))]; % elevation
    end
end
```

For non-additive noise: `z = measFcn(state, noise, mp)` — noise vector is injected.

## Custom Measurement Jacobian (for EKF/CKF)

If using EKF or CKF, provide the Jacobian:

```matlab
function J = myMeasJacFcn(state, mp)
    % J: M×N matrix where M = measurement dim, N = state dim
    % Compute analytically or use numerical differentiation:
    N = numel(state);
    M = numel(myMeasFcn(state, mp));
    J = zeros(M, N);
    dx = 1e-6;
    z0 = myMeasFcn(state, mp);
    for i = 1:N
        sp = state; sp(i) = sp(i) + dx;
        J(:,i) = (myMeasFcn(sp, mp) - z0) / dx;
    end
end
```

For UKF or particle filter, no Jacobian is needed.

## Custom Filter Initialization Function Template

```matlab
function filter = myFilterInit(detection)
    % Extract measurement and parameters
    meas = detection.Measurement;
    mp = detection.MeasurementParameters;
    noise = detection.MeasurementNoise;
    
    %% 1. Inverse measurement model: derive initial state from measurement
    switch mp.SensorType
        case 'tdoa'
            % TDOA gives a hyperboloid — cannot fully initialize position
            % Use a prior or range-parameterized approach
            state0 = [0; 0; 0; 0; 0; 0];  % poor init — needs fusion
            
        case 'bearing'
            % Angle-only: assume range, project to Cartesian
            assumedRange = 10000;  % m
            az = deg2rad(meas(1)); el = deg2rad(meas(2));
            localPos = assumedRange * [cosd(meas(2))*cosd(meas(1)); 
                                       cosd(meas(2))*sind(meas(1)); 
                                       sind(meas(2))];
            pos = mp.SensorRotation' * localPos + mp.SensorPosition;
            state0 = [pos(1); 0; pos(2); 0; pos(3); 0];
    end
    
    %% 2. Create filter with matching models
    filter = trackingEKF( ...
        StateTransitionFcn=@constvel, ...
        StateTransitionJacobianFcn=@constveljac, ...
        MeasurementFcn=@myMeasFcn, ...
        MeasurementJacobianFcn=@myMeasJacFcn, ...
        State=state0, ...
        StateCovariance=diag([1e6, 100, 1e6, 100, 1e6, 100]), ...
        ProcessNoise=diag([10, 10, 10]), ...
        HasAdditiveProcessNoise=false, ...
        HasAdditiveMeasurementNoise=true);
    
    %% 3. Set measurement noise from detection
    filter.MeasurementNoise = noise;
end
```

## Key Constraints

1. **Measurement function signature is fixed on the filter** — all sensors share it
2. **State vector must be consistent** — the motion model and measurement model must agree on state layout
3. **MeasurementParameters is the ONLY way to differentiate sensors** in multi-sensor setups
4. **Filter init must handle all SensorIndex values** if branching initialization by sensor

## Multi-Sensor Custom Setup

```matlab
function filter = myMultiSensorInit(detection)
    mp = detection.MeasurementParameters;
    meas = detection.Measurement;
    
    % Branch inverse model by sensor type
    switch mp.SensorType
        case 'radar'
            state0 = invertRadarMeas(meas, mp);
            cov0 = radarInitCovariance(meas, mp);
        case 'tdoa'
            state0 = invertTDOA(meas, mp);
            cov0 = tdoaInitCovariance();
    end
    
    % Same filter structure regardless of initializing sensor
    filter = trackingUKF( ...
        StateTransitionFcn=@constvel, ...
        MeasurementFcn=@myMeasFcn, ...
        State=state0, ...
        StateCovariance=cov0, ...
        ProcessNoise=diag([10, 10, 10]), ...
        HasAdditiveProcessNoise=false, ...
        HasAdditiveMeasurementNoise=true);
    filter.MeasurementNoise = detection.MeasurementNoise;
end
```

## Delivering to the User

Always deliver the custom measurement function, Jacobian (if EKF/CKF), and filter init as a matched set. Verify they work together:

```matlab
% Quick validation
det = objectDetection(0, testMeas, 'MeasurementParameters', mp, 'MeasurementNoise', testNoise);
filter = myFilterInit(det);
expectedMeas = myMeasFcn(filter.State, mp);
% expectedMeas should be close to testMeas
```

----

Copyright 2026 The MathWorks, Inc.
deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall