Learning-Based Control and Optimization

This project thread explores how learning-based methods can be integrated with classical control and optimization to design more capable engineering systems. The emphasis is on practical control design problems where robustness, stability, and performance must be balanced.

Core Questions

• When does reinforcement learning provide measurable advantages over classical tuning and optimization methods?
• How can learning-based methods be constrained so that they remain interpretable and engineering-relevant?
• How can optimization and control techniques transfer across domains such as vehicles, suspensions, and robotic systems?

Methods

• Reinforcement learning for controller design
• Classical control baselines including LQR, PID, and full-state feedback
• Multi-objective optimization
• Simulation-driven design and evaluation

Representative Outputs

• Reinforcement Learning-Based Optimal Control Design for Active Suspension Systems
• Robotic Optimal Path Planning via Multi-Objective Dolphin Echolocation Optimization (DEO)
• A Machine Learning Approach for Detecting Unauthorized Drone Operators

Research Stack

• MATLAB and Simulink
• Python
• Numerical optimization workflows
• Simulation environments for dynamic systems