SwarmSync

Distributed Model Predictive Control (DMPC) for Multi-Drone Motion Planning with Dynamic Goal Locations

SwarmSync extends traditional DMPC approaches to handle dynamic goal tracking in multi-agent drone systems using physics-based simulation with gym-pybullet-drones.

---

Overview

Traditional DMPC approaches assume static goal locations known at the start. SwarmSync addresses real-world scenarios where drones need to:

• Track moving targets in real-time
• Adapt trajectories to changing mission objectives
• Maintain collision-free paths with dynamic goals
• Coordinate multiple agents in dense environments

---

Technical Approach

Core Components

• Bézier Curve Parameterization: Smooth trajectory generation
• On-Demand Collision Avoidance: 50% faster than traditional BVC methods
• Event-Triggered Replanning: Adaptive response to disturbances and goal changes
• Distributed QP Optimization: Real-time solving at 5 Hz planning rate
• gym-pybullet-drones Integration: Physics-based quadrotor dynamics with Crazyflie 2.0 models

Key Innovation

Adapting the optimization problem to track moving targets while preserving collision-free guarantees in a realistic physics simulator.

---

Installation

Prerequisites

# Python 3.7+
pip install numpy matplotlib pybullet pybullet_data tqdm

Clone the Repository

git clone https://github.com/Shreyas0812/SwarmSync.git
cd SwarmSync

---

🎮 Usage

Visualize Trajectories with Static Goals

python viz_trajectory.py

Visualize Trajectories with Dynamic Goals

python viz_trajectory_goals.py --trajectory trajectories.txt --goals goals.txt

Or use the shorthand:

python viz_trajectory_goals.py -t trajectories.txt -g goals.txt

Plot Results

python plot_results_python.py

This generates:

• Distance to target over time
• 3D trajectory animations
• Performance metrics

---

Data Format

trajectories.txt

Contains the planned trajectories for all drones:

• Line 1: N N_cmd pmin[3] pmax[3] (metadata)
• Lines 2-4: Initial positions po for N drones (x, y, z)
• Lines 5-7: Final positions pf for N_cmd commanded drones
• Lines 8+: Trajectory data pk (3 lines per drone: x, y, z coordinates over time)

goals.txt

Contains dynamic goal trajectories:

• 3 lines per drone (x, y, z coordinates)
• Each line has timestep columns showing goal evolution

---

Project Structure

SwarmSync/
│
├── viz_trajectory_goals.py    # PyBullet visualization with dynamic goals
├── viz_trajectory.py           # PyBullet visualization with static goals
├── plot_results_python.py      # MATLAB-equivalent plotting script
│
├── trajectories.txt            # Generated trajectory data
├── goals.txt                   # Dynamic goal trajectories
│
├── DMPC-Pitch.md               # Project presentation (Marp slides)
├── DMPC-Pitch.pdf              # Compiled presentation
│
└── README.md                   # This file

Expected Performance (from base paper)

• ✓ 90%+ success rate with 30 agents in dense environments
• ✓ 50% reduction in travel time vs. traditional methods
• ✓ Real-time capable (20 Hz control with 20 drones)

---

Contributors

• Shreyas Raorane (@Shreyas0812)
• Kabir Puri (@kaRpuri)

---