| name | dept-matlab |
| description | Specialist Department: MATLAB & Simulation. Activated when project involves .m/.mlx/.slx files, Simulink models, Stateflow, code generation, fixed-point conversion, or model-based design. Trigger for "MATLAB", "Simulink", "Stateflow", "Embedded Coder", "MATLAB Coder", "fixed-point", "transfer function", "Bode plot", "state-space", "MIL", "SIL", "PIL", "HIL", "model-based design", "code generation", "designfilt", "fft", "ode45", or any MathWorks product or numerical computing context.
|
Specialist Department: MATLAB & Simulation
Activation
Activates when ANY of: .m, .mlx, .slx, .mdl files detected; MATLAB path setup
(startup.m, pathdef.m); toolbox usage; Simulink blocks; Stateflow charts; Embedded
Coder configuration; or user explicitly states MATLAB context.
Phase Injections Into All Roles
→ Requirements Analyst
Additional NFR categories to capture:
- Numerical accuracy: Tolerance bands for each output signal. Reference comparison method
(analytical solution, validated tool, high-fidelity simulation).
- Fixed-point requirements: Word length constraints, overflow strategy (saturate/wrap),
precision requirements per signal.
- Code generation targets: Target MCU, execution time budget per step, memory budget,
MISRA compliance for generated code.
- Simulation requirements: Solver type (fixed/variable step), step size, simulation
duration, real-time factor target.
- Toolbox dependencies: Which MATLAB toolboxes are required (Signal Processing, Control
System, Image Processing, etc.).
→ System Engineer
Additional design concerns:
- MATLAB package namespace: Organize code in
+pkgname/ packages to avoid name collisions.
- Simulink model hierarchy: Top model → subsystems. Bus naming conventions.
- Signal naming with units:
velocity_mps, temperature_degC, pressure_kPa, angle_rad.
Every signal name includes unit suffix.
- Parameter naming:
maxTorque_Nm, filterOrder, sampleRate_Hz. Include unit.
- Fixed-point scaling strategy: Document scaling for each signal path. Use Fixed-Point
Designer to analyze ranges and propose types.
- MIL/SIL/PIL/HIL test levels:
- MIL: Simulink model tested in Simulink.
- SIL: Generated C code tested on host.
- PIL: Generated code on target MCU.
- HIL: Generated code on target with simulated plant.
→ Backend Developer (MATLAB Developer)
MATLAB coding standards:
- Function headers: H1 line + full help block. Use
arguments block (R2019b+) for type
and range validation on all public functions.
function [output] = calculateTrajectory(pos, vel, dt, options)
%CALCULATETRAJECTORY Compute projectile trajectory with drag.
% [POS] = CALCULATETRAJECTORY(P0, V0, DT) computes trajectory...
arguments
pos (3,1) double
vel (3,1) double
dt (1,1) double {mustBePositive}
options.dragCoeff (1,1) double {mustBeNonnegative} = 0.47
end
- Vectorize: Prefer vectorized operations over loops. Use
arrayfun, bsxfun.
Document when loops are intentional (memory, algorithm requirement).
- Pre-allocate:
result = zeros(n, m) before loops. NEVER grow arrays in loops.
- No eval/evalin/feval(string): Security risk and prevents static analysis.
- No global variables: Pass data through arguments or object properties.
- eps for float comparison: NEVER
if (a == b). ALWAYS if abs(a-b) < tol.
- Use
\ not inv(): x = A \ b is numerically stable. x = inv(A) * b is not.
Check condition: if cond(A) > 1e12, warning('Ill-conditioned'); end
- Naming: camelCase functions, PascalCase classes, UPPER_SNAKE constants.
snake_case for Simulink models, camelCase signals with unit suffix.
- 80 char line max. Wrap with
.... 4-space indent. One statement per line.
- Section breaks:
%% Section Name to organize scripts.
- Simulink: Run Model Advisor before delivery. Pass MAAB / ISO 26262 / DO-331 checks
as applicable. Complete signal naming (no unlabeled wires).
→ Unit Tester
MATLAB testing standards:
- Use
matlab.unittest.TestCase with method-based tests.
- Traceability:
% Validates FR-001, AC-001-1 comment on every test.
- Numerical assertions:
testCase.verifyEqual(actual, expected, 'AbsTol', 1e-10) or
'RelTol', 1e-6. NEVER use exact equality for floating-point.
- Performance:
timeit(@() functionToTest(args)) to validate NFR timing budgets.
- Coverage: Use
CodeCoveragePlugin for coverage reporting.
- Test NaN/Inf: Verify explicit handling (not silent propagation).
- Test empty inputs:
[], {}, empty strings.
- Test ill-conditioned inputs: Near-singular matrices, near-zero denominators.
- Test dimensional correctness: Input/output sizes match specification.
- Simulink testing:
sltest.TestCase for model testing. Equivalence testing for
MIL vs SIL, SIL vs PIL. Use test harness for each subsystem.
→ Integration Tester
- MIL → SIL equivalence: Same inputs to Simulink model and generated C code → outputs
match within tolerance.
- SIL → PIL equivalence: Same inputs on host and target MCU → outputs match.
- Model-level integration: Test subsystem interactions within the top model.
- Hardware integration: If targeting real hardware, test sensor/actuator interfaces.
→ Security Engineer
- No
eval, evalc, evalin, feval(string) — injection risk.
- No
system() or ! command with user-supplied strings.
- Validate all file paths before
fopen, load, save.
- Use
matfile or datastore for large data (not load into workspace).
- Handle file I/O errors with try/catch.
- For production generated code: Sign code artifacts, verify authenticity.
→ Documenter
- Document all functions with full MATLAB help block format.
- Simulink models: Document each subsystem purpose in block annotations.
- Include signal flow descriptions for complex models.
- Data dictionary: Document all workspace variables, bus definitions, enumerations.
→ Release Engineer
Additional release artifacts:
- Code generation report (if Embedded Coder used): HTML report from codegen.
- Fixed-point comparison report: Float vs fixed-point results comparison with error analysis.
- MIL/SIL equivalence test report: Showing matching results within tolerance.
- Model Advisor results: Showing all checks passed or documented exceptions.
- Toolbox dependency list:
matlab.codetools.requiredFilesAndProducts output.
- MATLAB version and toolbox versions: For reproducibility.
Common MATLAB Toolboxes
| Toolbox | Use Case |
|---|
| Signal Processing | Filtering, FFT, spectral analysis |
| Control System | Transfer functions, Bode, root locus |
| Image Processing | Filtering, segmentation, feature extraction |
| Statistics & Machine Learning | Regression, classification, clustering |
| Optimization | Linear/nonlinear solvers, genetic algorithms |
| Symbolic Math | Analytical solutions, symbolic derivation |
| Simulink | Dynamic system modeling and simulation |
| Stateflow | State machines, sequential logic |
| Embedded Coder | Production C/C++ code generation |
| MATLAB Coder | General-purpose C/C++ code generation |
| Simulink Test | Test harness management, equivalence testing |
| Fixed-Point Designer | Fixed-point conversion and optimization |
| DSP System | Streaming signal processing |
| Communications | Modulation, coding, channel models |
| Aerospace | Flight dynamics, coordinate transforms |
| Robotics System | Kinematics, path planning, ROS interface |
| Automated Driving | Perception, planning, vehicle dynamics |
| Sensor Fusion & Tracking | Kalman filters, multi-object tracking |
| Deep Learning | Neural networks, training, inference |
| Computer Vision | Feature detection, object recognition |
| Radar | Waveform design, target detection |
Reference Files
references/numerical-computing.md — Float pitfalls, FFT patterns, filter design, control
systems, fixed-point notation, common error fixes.
