A production-ready, multi-rate MATLAB/Simulink architecture for GPS-denied UAV state estimation and robust flight control.
Traditional academic UAV simulations rely on idealized, continuous-time integrators, perfect sensors, and synchronous logic. When these algorithms are ported to embedded hardware, they fail catastrophically due to latency, asynchronous sampling, and sensor noise.
This repository bridges the gap between academic theory and embedded reality. It is an Enterprise-Grade Flight Stack built entirely on a multi-rate, discrete-time scheduling backbone. It enforces strict Interface Contracts (Simulink.Bus), utilizes official Aerospace Blockset models (ISA Atmospheres, Dryden Wind), and features a 15-state Extended Kalman Filter (EKF) capable of autonomous navigation in entirely GPS-denied environments.
This is not a toy model; it is a direct pathway to MISRA-C compliant firmware generation for flight controllers.
The system enforces strict modularity and separation of concerns across 5 primary domains:
- Mission & Guidance: Jerk-limited S-curve trajectory generation and Stateflow supervisory logic (Takeoff, Loiter, RTL).
- Control Allocation: Cascaded feedback loops (Position → Attitude → Body Rates) feeding an optimal mixing matrix with saturation protection.
- Physical Plant: 6-DOF rigid body kinematics (
aerolib6dofcg), non-linear thrust LUTs, asymmetric ESC delays, and aerodynamic drag. - Estimation (EKF): 15-state Extended Kalman Filter operating asynchronously. Continuously estimates pose and tracks IMU/Magnetometer bias random walks.
- Fault Detection (FDIR): Innovation Mahalanobis distance checks, motor desynchronization detectors, and brownout monitors.
- Multi-Rate Execution: Independent loop rates governed by a master task scheduler (e.g., EKF @ 200Hz, Control @ 400Hz).
- Strict I/O Contracts: All cross-domain data is strongly typed via predefined Signal Enumerations and Interface Buses.
- Advanced State Estimation: 15-DOF EKF with quaternion re-orthogonalization and divergence bounds.
- Aerospace Dynamics: Implementation of ground effect, propwash interaction, and mass/CG shifts.
- Embedded Coder Ready: Configured for direct MISRA-C/C++ compilation targeting ARM Cortex-M hardware.
- Automated CI/CD: GitHub Actions workflows for continuous regression testing and code generation memory budgeting.
- MATLAB R2023a / R2024a / R2025a / R2026a (or newer).
- Required Toolboxes:
- Aerospace Blockset
- Stateflow
- Simulink Control Design
- Embedded Coder (Optional, for firmware generation)
-
Clone the repository:
git clone https://github.com/ARYA-mgc/UAV_Advanced_Architecture.git cd UAV_Advanced_Architecture -
Initialize the Enterprise Parameter Space: Open MATLAB and run the master integration script. This automatically loads all Interface Contracts, Scenarios, and parameters into your workspace.
run('Runtime/init_all.m') -
Build and Launch the Simulink Model:
run('Codegen/build_ins_simulink_model.m') -
Verify Stability (Linearization):
run('Validation/trim_and_linearize.m')
The repository adheres to a strict NASA-style hierarchical layout containing over 50 isolated scripts:
Interfaces/— StrictSimulink.Buscontracts, signal enums, and unit registries.Runtime/— Task schedulers, asynchronous clock managers, and bus latency injectors.Guidance/— Mission parsers, 3D waypoint managers, and dynamic replanning logic.Control/— Optimal LQR synthesis, PID schedules, and anti-windup logic.Plant/— Motor dynamics, battery sag curves, and 6-DOF environment interactions.Sensors/— IMU noise seeds, ADC quantization limits, and GPS dropout models.Estimator/— Core EKF algorithms, covariance tuning, and dead reckoning logic.Faults/— Supervisory fallback matrices, thrust loss detectors, and glitch monitors.Scenarios/— Isolated programmatic flight tests (e.g., severe wind gusts).Validation/— Monte Carlo stress campaigns, linearization tools, and unit tests.Logs/— CSV/ULog exporters and telemetry replay pipelines.Codegen/— Polyspace/MISRA configurations and stack usage estimators.
The /Scenarios directory provides programmatic flight test harnesses to stress the system:
- Aggressive Maneuvers (
aggressive_turn_test.m): Validates the feedforward control and rate limits. - Environmental Stress (
hover_wind_gust.m): Injects sudden Dryden turbulence while monitoring EKF position hold. - Hardware Degradation (
motor_failure_takeoff.m): Simulates a 30% thrust loss on Motor #2; verifies the control allocator and FDIR logic successfully trigger a degraded RTL.
When subjected to standard continuous validation runs, the flight stack achieves:
- Hover Stability (RMSE): Position < 0.2m | Attitude < 1.5° (in 15 knot crosswinds).
- State Estimation Drift: < 2 meters of accumulated drift over 60 seconds during total GPS-denial.
- Actuator Latency Resilience: Stable up to 80ms of artificially injected ESC/CAN-bus latency.
Rigorous testing is baked into the architecture:
- Monte Carlo Campaigns: The
run_monte_carlo.mscript programmatically iterates through randomized permutations of mass (±20%), CG offsets, and wind states to prove broad-spectrum controller stability. - Trim & Linearization: Automated equilibrium extraction ensures mathematically proven gain and phase margins.
- Continuous Integration: The
.github/workflows/sim_regression.ymlexecutes headless testing on every pull request.
The system operates under the assumption that hardware will fail.
- Sensor Rejection: The EKF utilizes continuous Mahalanobis distance gating to ignore sudden spikes in Barometer or Magnetometer data.
- Supervisory Fallbacks: If thrust saturation exceeds predefined timeouts, the Stateflow engine bypasses the active mission plan and forces an Emergency Landing sequence.
- Phase 1: Integration of Computer Vision (Optical Flow / VIO) into the Estimator layer.
- Phase 2: Real-time Model Predictive Control (MPC) optimization for agile obstacle avoidance.
- Phase 3: Hardware-In-The-Loop (HIL) deployment onto PX4 / STM32 architectures.
ARYA MGC
Lead Architect | Aerospace / ECE
Contributions and pull requests are strictly reviewed against the existing Interface Contracts and require passing CI unit tests.
This project is licensed under the MIT License - see the LICENSE file for details. Academic and commercial utilization is permitted provided appropriate attribution is given. For proprietary integration support, please refer to the contact information.