UAV Model Predictive Control (MPC) — Path Correction

This project implements a full Model Predictive Control (MPC) pipeline for 3D quadrotor (UAV) path correction. It simulates position tracking and path correction in Python, then transitions seamlessly to real-world integration with the DJI Tello EDU drone.

---

🚀 Features

• 3D double integrator model (x, y, z position and velocity)
• Real-time MPC optimization using CasADi and IPOPT
• Visualization of:
  • Position tracking vs. reference
  • XY top-down trajectory
  • Control input accelerations
  • Tracking error
• Modular design to connect with real drones
  • Swappable reference trajectories (circle, figure-8, custom)
  • Vision-based position estimation via OpenCV
  • Live MPC → velocity control mapping for DJI Tello EDU

---

🛠️ Dependencies

pip install numpy matplotlib casadi opencv-python djitellopy

---

🧠 How It Works

The MPC solves a finite-horizon optimization problem at each timestep to minimize the difference between the drone’s actual and desired positions while respecting acceleration limits.

The real-time version reads visual feedback (e.g., from a face or AprilTag) to estimate the drone’s relative position and continuously adjusts the drone’s velocity to keep the target centered and at a constant distance.

---

🧪 Run Simulation

python -m src.sim.run_mpc