Roadmap
The current release is v0.3.0, a bounded offline engineering workbench. The README and the capability registry describe what exists; the operating guide describes its limits. Future work below is not part of this release and has no promised delivery date.
The rule for each milestone remains that a reader can build that version and run a complete, reviewable study. Completing a workflow does not qualify a tool or validate a new aircraft model.
v0.1 — trim and linearisation
Implemented: frames and units, ISA atmosphere, quaternion rigid-body dynamics, fixed-step RK4, the CLI and YAML runner, the local derivative aircraft model, straight-line trim, finite-difference linearisation and classified aircraft modes. Published-reference comparisons for the NT-33A condition and the underlying numerical methods are in VERIFICATION.md.
The known phugoid discrepancy and its labelled regression locks remain documented in the investigation note. They are not reclassified as validation successes by later releases.
v0.2 — frequency-domain analysis
Implemented: frequency response, gain/phase/delay crossovers, SISO disk margin, MIMO principal gains, sensitivity and complementary-sensitivity peak estimates. The published NT-33A HTML report adds plots generated from its computed run.
These original frequency-peak searches use a finite refined grid. A sampled peak can underestimate the true norm, making a derived margin optimistic. Their reference tests do not establish a bound for every possible transfer function. The original capabilities retain this limitation and their identity.
v0.3 — bounded offline workbench
Implemented scope:
- Continuous CARE and LQR synthesis with residual, conditioning and stability evidence; filtered PID from explicit gains; channel selection, series and negative-feedback interconnections.
- Hamiltonian H-infinity norm brackets, S/T norm brackets and SISO disk-size bounds. These supplement the sampled analyses and reject unresolved numerical cases. They are floating-point numerical bounds, not interval proofs.
- Fixed-step linear response and local nonlinear aircraft simulation with explicit position bounds, rate limits and positive first-order lags for all four controls. Continuous state feedback and constant command increments are supported about a level translating reference.
- A complete NT-33A local design example, trajectory CSV, Markdown and self-contained HTML tables, strict input schemas, contained output writes, input snapshots and immutable content-addressed run manifests.
- Installable C++ libraries and CLI, dependency notices, installed-consumer and release-archive checks, and expanded numerical and integration tests.
Evidence for the CARE solver is distinguished from aircraft/controller validation in the synthesis design record. The example's costs and actuator specifications are illustrative. The model still describes a single local aerodynamic reference condition.
Earlier roadmap versions assigned v0.3 to a desktop application. That scope is deferred from the release baseline. The subsequent experimental M2 increment adds a native macOS editor for scalar projects and typed linear aircraft/controller study imports, with a Dim theme, native block/sample tables and saved-revision review/restore. Source matrices and diagnostics remain attached after edits. It has no 3-D viewport or signed desktop installer. It does not complete the former v1.0 feature list.
Future work — not implemented
The following are separate extensions, each requiring its own model assumptions, reference data and acceptance criteria before it becomes a release commitment.
| Area | Remaining work |
|---|---|
| Aircraft fidelity | Additional independently validated conditions and aircraft, scheduled aerodynamic data, propulsion and configuration models, quantified model uncertainty |
| Control design | Discrete-time control, sensors and estimators, explicit delay models, constrained synthesis, gain scheduling and robust-design methods |
| Assessment | Handling-quality criteria, root locus, uncertainty campaigns and application-specific acceptance reports |
| Simulation and identification | Disturbances, measured-data identification, sampled control execution, SITL/HIL interfaces and independently measured hardware responses |
| User interface | Desktop shell, interactive plots and pipeline editor, engineering 3-D visualisation, platform installers |
| Extensions and automation | Stable C plugin ABI, aerodynamic/sensor extension contracts, AI and MCP interfaces with separate execution controls and evaluations |
A future stable release needs an explicit scope and compatibility contract; it cannot be inferred from the number of completed capabilities. There is no onboard deployment or tool-qualification claim in this roadmap.
Reference data and evidence
Shipped numerical data requires documented source and rights provenance under ADR-0007. A prospective dataset whose redistribution rights are unresolved does not become acceptable merely because it would improve an example. Where no published comparison is available, the capability remains labelled unvalidated in the verification report; analytic tests and regression locks retain their separate purposes.
Proposed end-product completion plan
The product plan develops the future work into a proposed offline desktop engineering product with Simulink-style executable block modeling. It links a prioritized feature catalog, discovery investigations, milestone action plans and assurance gates. These are planning documents, not newly implemented capabilities or a qualification claim; the current release scope above remains unchanged.