Drone Multi-Control Algorithm for Formation Control

This project focuses on developing and optimizing formation control algorithms for multi-UAV systems. Using distance-based formation and gradient descent optimization, our approach enables drones to maintain safe distances, avoid collisions, and achieve precise formations.

Note

This is an optimization of the original gym-pybullet-drones/examples/pid.py repository , designed for compatibility with PyBullet

Classical Formation Control with Collision

Implemented Optimization with No Collision

Dependencies

This project tests its simulations on: PyBullet and uses Python 3.10+

For more information, please visit PyBullet's repository.

• PyBullet

Virtual Environment Setup

1. Install necessary packages:

   pip install pybullet
   pip intall numpy
   pip install matplotlib

2. Important Notice

   If using Windows 11, you may need to install Microsoft's Visual Studio and disable Windows 11 SDK from the Installation Details section to get the PyBullet simulator to work.

How to Use the Code

1. Using a python terminal

   python pid_square_form.py

   Simulation Variables:

   • H_STEP (the step-size for gradient descent): 0.15
   • DEFAULT_DURATION_SEC: 10
   • DEFAULT_NUM_DRONES: 4

2. Output

   1. Initializes the PyBullet simulation environment and sets up drones at random starting positions.
      • Can use np.random.seed(0) to get both in the same initial positions
   2. Calculates desired inter-drone distances based on a target square formation.
   3. Runs the formation control algorithm using gradient descent to minimize formation error.
   4. Applies a repulsive force for collision avoidance when drones are closer than the safe distance for the first run function.
   5. Updates drone positions in real time using PID controllers to track gradient-based target points.
   6. Displays a live 3D simulation in the PyBullet viewer showing drones moving into formation.

Example Plots

1. Gradient magnitude over time for each drone

2. Objective function error over time comparison