Simulink Requirements

MATLAB Requirements Toolbox — Requirements & Traceability

This skill covers everything in the slreq API: creating and reading requirements, managing traceability links, checking verification coverage, allocating requirements to architecture components, and auditing link health.

For architecture phases (System Composer models, functional decomposition, functional→physical allocation) see the mbse-architecture skill.

See references/api-quickref.md in this skill folder for a compact one-page API reference.

The Two File Types

When a .slmx file is created

slreq writes a .slmx file only when its companion artifact is the source of at least one link. The LinkSet is keyed on the source artifact, not the destination.

• MyReqs~slreqx.slmx appears only if some link's source is a requirement in MyReqs.slreqx (e.g., slreq.createLink(srcReq, destReq) where srcReq lives in MyReqs)
• MyModel~mdl.slmx appears only if some link's source is a model element in MyModel.slx (typical: slreq.createLink(component, req) with Type Implement — the component is the source)

An artifact that is only ever a link destination never gets a paired .slmx. In a typical MBSE project:

Reporting rules for build scripts

Because .slmx files are conditional, never claim a .slmx was produced based on the API calls you made — always verify with isfile before reporting. A script that creates only SN→SR Derive links will produce StakeholderNeeds~slreqx.slmx but not SystemRequirements~slreqx.slmx, even though both .slreqx files exist.

Idempotent cleanup and project registration for .slmx files must both guard with isfile:

if isfile(snLinks), delete(snLinks); end   % cleanup — safe whether or not it exists

The registerWithProject helper already does this — it skips files that don't exist on disk — so passing a non-existent .slmx path is a no-op, not an error.

Requirements (Phases 1–2)

Two-Level Structure

Each SR traces back to one or more SNs via a Derive link.

Well-Formed Shall-Statements

• One obligation per requirement ("shall", not "should" or "will")
• Measurable and testable — include numeric criteria where possible
• Avoid < and > in Description fields — the Requirements Editor treats them as HTML. Use "not exceeding", "at least", "greater than", etc.

Good: The system shall respond with latency not exceeding 100 ms. Avoid: The system shall respond with latency < 100 ms.

Creating Requirement Sets

slreq.clear();
if isfile('MyReqs.slreqx'), delete('MyReqs.slreqx'); end
rs = slreq.new('MyReqs.slreqx');        % NOT slreq.createReqSet (does not exist)

req = rs.add();
req.Id          = 'SR-SYS-001';
req.Summary     = 'Short title';
req.Description = 'The system shall ...';
req.Rationale   = 'Why this requirement exists.';

rs.save();

Find a requirement by ID

req = rs.find('Id', 'SR-SYS-001');

Valid Link Types

Creating Links

% Req-to-req derivation: parent (e.g. SN) decomposes into derived child (e.g. SR)
lnk = slreq.createLink(parentReq, childReq);
lnk.Type = 'Derive';

% Model block to req (model must be open in Simulink)
lnk = slreq.createLink(blockHandle, req);
lnk.Type = 'Implement';

% Test case requirement to SR
lnk = slreq.createLink(tc, sr);
lnk.Type = 'Verify';

slreq.saveAll();   % always call after creating cross-artifact links

Side effect: {modelName}~mdl.slmx link store

The first time you create a link whose source is an element of a Simulink/System Composer model (component, subsystem, or block), slreq writes a {modelName}~mdl.slmx file next to the .slx. This is the normal case for Implement links in this workflow — slreq.createLink(component, req) has the component as source, so the link lives in the model's LinkSet, not the requirement set's.

A model that only ever receives links (no element of it is used as a link source) does not get a ~mdl.slmx. In MBSE practice this is rare: architecture models almost always have Implement links where they are the source.

After creating links whose source is in a model, register both the .slx and the .slmx with the project, guarding .slmx registration with isfile since the file may not yet exist:

addFile(proj, fullfile(archDir, 'MyModel.slx'));
slmx = fullfile(archDir, 'MyModel~mdl.slmx');
if isfile(slmx), addFile(proj, slmx); end

Forgetting to register an existing .slmx causes project file-system checks to fail and the traceability links won't travel when the project is shared.

Requirements Script Skeleton

See code/buildMyRequirements.m for the full parameterized function:

buildMyRequirements(snFile, srFile)

Exporting a Requirement Set to Excel

There is no public slreq Excel-export API. slreq.export emits ReqIF only, and the Requirements Editor's File → Export → Microsoft Excel is GUI-only. To script an xlsx export, build the table yourself with writetable.

Requirement sets can be hierarchical. find(rs, 'Type', 'Requirement') returns all requirements flat in storage order, which silently drops the parent/child structure the Requirements Editor displays. To preserve hierarchy:

• Find top-level reqs by filtering for ~isa(r.parent(), 'slreq.Requirement') — top-level items have the ReqSet as their parent, not another requirement
• Recurse via the children() method (a method, not a property)
• Emit rows depth-first so each parent precedes its children
• Sort siblings with a natural sort on the Index string ("1", "1.2", "1.10" as numeric tuples — lexicographic string sort would put 1.10 before 1.2)
• Write Index, Depth, ParentIndex columns so the hierarchy is recoverable from the xlsx alone

Note also that r.Id may be an auto-assigned SID like #13 if the set was authored in the Editor without user IDs — always include the Index column as a stable user-meaningful identifier.

Extract parent IDs from incoming Derive links so the DerivedFrom column is populated instead of stuffing parent refs into Rationale:

function ids = deriveParents(req)
    ids  = strings(0,1);
    lnks = req.inLinks();
    for k = 1:numel(lnks)
        if strcmp(lnks(k).Type, 'Derive')
            % getSourceLabel() returns "ID Summary"; strtok pulls just the ID
            ids(end+1,1) = string(strtok(lnks(k).getSourceLabel())); %#ok<AGROW>
        end
    end
end

See code/exportRequirementsToExcel.m for the full parameterized function:

exportRequirementsToExcel(slreqxFile)

Importing a Requirement Set from Excel

slreq.import handles Excel natively, but has several gotchas worth wrapping: it treats the header row as a requirement (use rows=[2 lastRow] to skip it), it auto-creates a Container node wrapping the items, and — most importantly — it does not save to disk (the returned ReqSet is marked Dirty=1; call .save() explicitly).

Use AsReference=false to get an editable copy rather than read-only references to the xlsx, and map columns with idColumn / summaryColumn / descriptionColumn / rationaleColumn.

Flatten the auto-created Container by default

slreq.import always wraps imported items under a single Container node named "<File>!<Sheet>", which becomes index 1 in the set and pushes every actual requirement to 1.1, 1.2, etc. Users viewing the set in the Requirements Editor see a useless extra hierarchy level. The wrapper helper unwraps this by default: it forward-promotes each direct child of the Container to top level, then removes the empty Container. Real nested hierarchy (e.g. when importing ReqIF with actual parent/child structure) is preserved — the helper never touches children of non-Container nodes.

Two invariants to preserve if editing the flatten logic:

• Exactly one Container at top level — flatten only the auto-import wrapper. If a set has zero or multiple top-level Containers, leave them all alone (that's not the wrapper signature).
• Forward-promote, not reverse — promoting children in reverse reverses sibling order in the resulting flat list. Iterate 1:numel(kids).

Helper

See code/importMyRequirements.m for a wrapper that applies the defaults and saves:

importMyRequirements(xlsxFile, setName)
importMyRequirements(xlsxFile, setName, flatten=false)  % keep the Container

Verification (Phase 7) — TC Requirements

Test cases live in their own requirement set (TestCases.slreqx), separate from system requirements. Each TC is an slreq.Requirement linked to its SR with a Verify link.

SR-SYS-001  ←[Verify]─  TC-SYS-001

Each TC captures a test in prose — setup, stimulus, pass criterion — and traces back to exactly one SR. This is the only verification layer in the workflow; there is no separate executable-test artifact.

TC Requirement Fields

A good description answers: Setup (initial conditions), Action (stimulus applied), Pass criterion (measurable result that constitutes success).

Test Case Script Skeleton

See code/buildMyTestCases.m for the full parameterized function:

buildMyTestCases(srFile, tcFile)

Verification Coverage Report

allSRs  = srSet.find('Type', 'Requirement');
covered = 0;
for i = 1:numel(allSRs)
    in_   = allSRs(i).inLinks();     % method on the req object — NOT slreq.inLinks()
    hasTc = false;
    for k = 1:numel(in_)
        if strcmp(in_(k).Type, 'Verify')
            hasTc = true;
            break;
        end
    end
    if hasTc
        covered = covered + 1;
    else
        fprintf('NOT COVERED: %s\n', allSRs(i).Id);
    end
end
fprintf('Coverage: %d / %d (%.0f%%)\n', covered, numel(allSRs), ...
    100 * covered / numel(allSRs));

Reading and Tracing Links

Loading Files for Analysis

To load a single file:

rs = slreq.load('path/to/MyReqs.slreqx');

To load all requirement and link files across a project tree, see code/loadProjectRequirements.m:

loadProjectRequirements(projRoot)

slreq.load() is for scripted analysis (no UI). slreq.open() opens the Requirements Editor UI — avoid it in analysis scripts.

File discovery note: proj.Files only lists top-level project files. Sub-project files don't appear. Always use dir(fullfile(root,'**','*.slreqx')) to find all files.

Querying Loaded Objects

% All loaded ReqSets and LinkSets
allReqSets  = slreq.find('type', 'ReqSet');
allLinkSets = slreq.find('type', 'LinkSet');

% Find a specific ReqSet by name
rs = slreq.find('type', 'ReqSet', 'Name', 'SystemRequirements');

% All requirements in a set (returns ALL node types: Functional, Container, etc.)
reqs = rs.find('Type', 'Requirement');

% Find by ID or SID
r = rs.find('Id', 'SR-SYS-001');
r = rs.find('SID', 5);

rs.find('Type','Requirement') returns every node — both Functional and Container types. Filter by r.Type if you only want leaf requirements.

Requirement Node Properties

r.Id              % user-assigned ID (string), e.g. 'SR-SYS-001'
r.SID             % internal integer, unique within the file
r.Type            % 'Functional', 'Container', 'Safety', 'Informational'
r.Summary         % one-line summary
r.Description     % HTML string — use getDescriptionAsText() for plain text
r.Rationale       % plain text rationale
r.Index           % hierarchical index, e.g. '2.1.3'

% Clean description (strips HTML formatting)
plainText = r.getDescriptionAsText();

% Navigate hierarchy
kids   = r.children();   % slreq.Requirement array of child nodes
parent = r.parent();     % slreq.Requirement or slreq.ReqSet (if top-level)

outLinks and inLinks

Every requirement has two directions of links:

out = r.outLinks();   % slreq.Link array
in_ = r.inLinks();    % slreq.Link array

% IMPORTANT: cannot vertcat outLinks/inLinks directly — iterate with index
for k = 1:numel(out)
    lnk = out(k);
    fprintf('  -> %s "%s"\n', lnk.Type, lnk.getDestinationLabel());
end

Reading Link Data

These methods work even when isResolved() is false:

lnk.Type                    % 'Derive', 'Implement', 'Verify', 'Relate', 'Refine'
lnk.getSourceLabel()        % human-readable label for the source artifact
lnk.getDestinationLabel()   % human-readable label for the destination artifact

% Resolution status
lnk.isResolved()            % true only if BOTH ends are resolved — often false, do not rely on it
lnk.isResolvedSource()      % source artifact is loaded
lnk.isResolvedDestination() % destination artifact is loaded

% Raw reference struct — ALWAYS readable, even when unresolved
src = lnk.source();           % struct: .domain, .artifact, .id  (source end)
ref = lnk.getReferenceInfo(); % struct: .domain, .artifact, .id  (destination end)

Link Domain Types

The .domain field identifies what kind of artifact the link points to:

Resolving a Req-to-Req Link Destination

ref = lnk.getReferenceInfo();
if strcmp(ref.domain, 'linktype_rmi_slreq')
    sid = str2double(ref.id);
    allRS = slreq.find('type', 'ReqSet');
    for i = 1:numel(allRS)
        [~, fn]      = fileparts(allRS(i).Filename);
        [~, artName] = fileparts(ref.artifact);
        if strcmpi(fn, artName)
            destReq = allRS(i).find('SID', sid);
            break;
        end
    end
end

Link Direction Semantics

SN  ──[Derive]──>  SR    (parent has outLink; derived child has inLink)
        Component  ──[Implement]──>  SR    (component has outLink; req has inLink)
        Test case  ──[Verify]────>  SR    (test has outLink; req has inLink)

In this workflow link direction goes parent/active → child/requirement-end for all three types — the parent SN points at the SR it decomposes into, the architecture element points at the SR it implements, and the test case points at the SR it verifies.

A requirement derives from another (parent) when it has inLinks() of type Derive. A requirement is implemented by an architecture element when it has inLinks() of type Implement whose source is a System Composer component (or a Simulink block). A requirement is verified by a test case when it has inLinks() of type Verify.

LinkSet Methods

links = ls.getLinks();                     % all links in a LinkSet
[broken, details] = ls.getBrokenLinks();   % links whose destination is gone
orphans = ls.getOrphanLinks();             % links whose source artifact is gone

ls.Artifact   % full path to the artifact this LinkSet belongs to
ls.Filename   % full path to the .slmx file itself

Coverage Analysis — Per-Requirement Status

See code/reportCoverageByReq.m:

reportCoverageByReq(rs)

Coverage Analysis — Aggregate Across All ReqSets

See code/reportCoverageAggregate.m — loads all files via loadProjectRequirements, then prints a per-ReqSet summary table:

reportCoverageAggregate(projRoot)

Link Health Report

See code/reportLinkHealth.m — operates on all currently loaded LinkSets; call loadProjectRequirements first:

reportLinkHealth()

Full Traceability Chain Trace

See code/traceRequirement.m:

traceRequirement(rs, reqId)

Common Pitfalls

slreq.inLinks(req) / slreq.outLinks(req) do not exist — these are methods on the requirement object: req.inLinks() and req.outLinks().

slreq.open() in scripts — use slreq.load() for scripted analysis. slreq.open() launches the Requirements Editor UI.

isResolved() is almost always false for model links — this is normal. Simulink block SIDs can't be resolved without opening the model. Always use getSourceLabel(), getDestinationLabel(), and getReferenceInfo() instead.

outLinks()/inLinks() arrays cannot be vertcat'd — iterate with index, or collect into a cell array first.

rs.find('Type','Requirement') returns Containers too — the Type property on returned objects is 'Container', 'Functional', etc. Filter if needed: reqs(strcmp({reqs.Type}, 'Functional')).

Description contains HTML — use r.getDescriptionAsText() to get clean text.

getImplementationStatus() requires update first — call rs.updateImplementationStatus() before calling r.getImplementationStatus(), otherwise it throws.

Delete .slmx link files alongside .slreqx files when rebuilding requirement sets. Stale .slmx files store cross-artifact links and will auto-open old model files on load, causing conflicts.

Linking a System Composer Interaction (sequence diagram) to a requirement doesn't persist on R2025b. Two dead ends:

• slreq.createLink(interactionObj, req) — with a raw systemcomposer.interaction.Interaction — throws a bare "Link creation failed" error.
• The struct workaround struct('domain','linktype_sc_interaction', 'artifact',modelName,'id',char(diagram.UUID)) does create a link in memory (and req.inLinks() shows it), but on next slreq.load the slreq.data.ReqData/loadLinkSet listener errors with "Artifact type mismatch: Expected linktype_rmi_simulink, Found linktype_sc_interaction" whenever the same .slx artifact already stores linktype_rmi_simulink Implement links from component→req wiring. slreq refuses to coexist two different artifact-type LinkSets on one .slx; the interaction-sourced link is silently dropped on reload.

Workarounds:

• Convention-based trace. Give the interaction a clear name and refer to it by name from requirements' Description or Rationale, or from a TC. Rely on humans to walk the trace.
• Companion TC. Create a dedicated TC-XXX-YYY whose description is "Execute the <interactionName> sequence on the model and verify message ordering and guard expressions." Verify link that TC to the target SR; slreq supports TC→SR Verify links normally.

Revisit in a future MATLAB release — mixed-domain LinkSets on one .slx may eventually be permitted, at which point the UUID-struct approach can be re-enabled. See the system-composer skill's Sequence Diagrams section for the programmatic API.

Don't stuff parent-reference text into Rationale. Writing "Derived from SN-SYS-001." into req.Rationale shadows the real purpose of the field AND collides with the DerivedFrom column on xlsx export. Keep Rationale for the why (what constraint or judgment picked this value); let the Derive link carry the parent reference, and extract it from inLinks() at export time.

Getting a parent requirement's Id from a link. lnk.source() returns a struct with .domain, .artifact, .id — but .id is the numeric SID, not the user-facing Id like 'SN-SYS-001'. For the Id string, use strtok(lnk.getSourceLabel()) — the label format is "ID Summary", so strtok pulls just the Id.

slreq.import does not save to disk. The function returns [refCount, reqSetFilePath, reqSetObj] — the path is where the file will live, but the ReqSet is only in memory and marked Dirty=1. Call reqSetObj.save() (or slreq.saveAll()) before closing, or the imported set vanishes. Observed with AsReference=false on Excel import; worth checking before relying on the default behavior in other modes.

Excel import treats the header row as a requirement unless you set rows=[firstDataRow lastDataRow]. Also auto-creates a Container node named "<File>!<Sheet>" wrapping the imported items; the importMyRequirements helper flattens it by default (see above), otherwise filter it out with r.Type == "Container" when iterating.

req.inLinks() auto-loads referenced ReqSets. Calling inLinks() on a requirement causes slreq to resolve link sources, which loads any .slmx / .slreqx files that reference this set. So a plain slreq.load(SystemRequirements.slreqx); req.inLinks(); may silently bring TestCases.slreqx into memory too (if any Verify link points here). This causes hidden state leakage across phases of a pipeline — if one phase loaded SR and walked its inLinks, the next phase's slreq.clear() must happen before that phase's slreq.load(), not after, and you can't assume slreq.find only returns sets you explicitly loaded.

req.remove() leaves orphan outLinks in the .slmx. Removing a requirement drops the requirement object, but its outgoing links (e.g. Verify links pointing at an SR) stay in the LinkSet as "unresolved source" entries until the LinkSet is reloaded. Symptom on next open: Warning: LinkSet for MyFile.slreqx contains N links with unresolved sources. Fix: clear the LinkSet explicitly before removing requirements. Pattern:

lnkSets = slreq.find('type','LinkSet','Artifact', reqFile);
for i = 1:numel(lnkSets)
    links = lnkSets(i).getLinks();
    for j = 1:numel(links), links(j).remove(); end
end
existing = rs.find('Type','Requirement');
for k = numel(existing):-1:1, existing(k).remove(); end

Idempotency pattern: load-or-clear-and-repopulate beats delete-and-new. slreq.new(file) intermittently fails with Can not create Requirement Set named X because of name conflict with X.slreqx even after slreq.clear() and a successful delete(file) — especially in long pipelines where earlier phases have touched the file. The robust idempotent pattern is:

if isfile(reqFile)
    rs = slreq.load(reqFile);
    clearLinkSet(rs);       % see above — avoids orphan-link warnings
    clearRequirements(rs);
else
    rs = slreq.new(reqFile);
end
% ... populate fresh ...
rs.save();

This is the pattern in Phase 9 of mbse-workflow/SKILL.md.