Optimization of LQR Controller Gains using Metaheuristic Algorithms

This repository presents the implementation of various metaheuristic optimization techniques to tune the Linear Quadratic Regulator (LQR) controller gains for feedback-linearized cruise missiles. The goal is to enhance control performance under practical constraints by automating the gain selection process.

---

Project Overview

Traditional LQR control design depends on manually selected weight matrices $Q$ and $R$, which can be suboptimal for real-world systems with nonlinearities, delays, and disturbances. In this project, the following global optimization algorithms are applied to optimize LQR performance:

• Genetic Algorithm (GA)
• Particle Swarm Optimization (PSO)
• Simulated Annealing (SA)

These methods are tested on a longitudinal flight dynamics model with Dryden turbulence, simulated using a custom Simulink model.

---

Repository Structure

├── Codes/
│   ├── Genetic_Algorithm.m           # GA implementation for LQR tuning
│   ├── Particle_Swarm.m              # PSO-based LQR optimization
│   ├── Simulated_annealing.m         # SA algorithm implementation
│   ├── Optimization_LQR.m            # LQR setup, fitness eval, call optimizer
│   └── drydenmodel.m                 # Turbulence model for aircraft simulation

├── Simulink Files/
│   └── Final_PAPER2_wtdelay_LQR.slx  # Full nonlinear aircraft model with LQR + delay

├── README.md                         # Project description and instructions

---

Setup & Execution

Software Requirements

• MATLAB R2021a or later
• Control System Toolbox
• Simulink

Steps to Run

1. Open Optimization_LQR.m to select the algorithm and launch optimization.
2. Run Genetic_Algorithm.m, Particle_Swarm.m, or Simulated_annealing.m independently for individual tests.
3. Simulate the optimized gains in Final_PAPER2_wtdelay_LQR.slx.
4. Use drydenmodel.m to inject atmospheric turbulence into the model.

---

Objectives & Contributions

• Compare different optimization algorithms for LQR gain tuning.
• Evaluate robustness under Dryden wind gust model.
• Automate weight tuning using population-based search methods.
• Validate control performance in a realistic missile dynamics environment.

---

Documentation

For complete theoretical background, design equations, training pipeline, and result analysis, refer to:

📄 BioinspiredLQR_preprint.pdf

Expected Outcomes

• Reduced settling time and overshoot
• Improved disturbance rejection
• Optimized control effort with better energy efficiency

---

Future Work

• Extend to MIMO systems and full 6-DOF UAV models.
• Integrate Deep Reinforcement Learning for adaptive LQR tuning.
• Implement a GUI-based controller synthesis tool.

---

License

This repository is released under the MIT License, allowing academic and research use.