Drone Show & Swarm Coordination

# ═══ MDS SITL DEMO — 10 DRONES IN SIMULATION ═══ # 1. Clone the repo git clone https://github.com/alireza787b/mavsdk_drone_show.git cd mavsdk_drone_show # 2. Install Python deps (3.11-3.13) pip install -r requirements.txt # 3. Run the SITL demo (starts 10 PX4 instances in Docker) # See: docs/guides/sitl-comprehensive.md for full walkthrough python3 multiple_sitl/multi_sitl_runner.py --num-drones 10 # 4. Start the GCS server (React dashboard) cd gcs-server && python3 app.py # 5. Open browser: http://localhost:5000 # You'll see 10 drones on a map, ready to fly formation # 6. Load a shape file (CSV choreography) # shapes_sitl/ has pre-made formations # Or design your own in Blender + SkyBrush # 7. Hit "Start Show" in the dashboard # Drones take off, fly to formation positions, execute choreographyMDS SITL

# ═══ SHAPE CSV FORMAT (config.csv) ═══ # Each drone has its own row. Columns define position offsets from origin. # hw_id, pos_id, x, y, z, ... # hw_id: Hardware ID (matches physical drone) # x: North offset (meters from formation origin) # y: East offset (meters) # z: altitude (positive = up) # Example: 5-drone line formation at 10m altitude 1, 1, 0.0, 0.0, 10.0 2, 2, 0.0, 3.0, 10.0 3, 3, 0.0, 6.0, 10.0 4, 4, 0.0, 9.0, 10.0 5, 5, 0.0, 12.0, 10.0 # For animated shows, use time-stamped waypoints per drone: # shapes/ folder contains CSV files with columns: # timestamp, north, east, down, red, green, blueCSV Format

# ═══ GENERATE CHOREOGRAPHY WITH PYTHON ═══ import numpy as np import csv num_drones = 20 radius = 15 # meters # Circle formation with open('shapes/circle_20.csv', 'w') as f: writer = csv.writer(f) for i in range(num_drones): angle = 2 * np.pi * i / num_drones north = radius * np.cos(angle) east = radius * np.sin(angle) alt = 20.0 # RGB: rainbow colors around the circle r = int(127 + 127 * np.sin(angle)) g = int(127 + 127 * np.sin(angle + 2.094)) b = int(127 + 127 * np.sin(angle + 4.189)) writer.writerow([0, north, east, -alt, r, g, b]) # For animated shows: multiple rows per drone with timestamps # t=0: circle, t=5: expand, t=10: star, t=15: landPython

# ═══ MDS SMART SWARM — LEADER-FOLLOWER ═══ # smart_swarm_src/ contains the swarm logic # Each drone runs drone_show.py which checks its role: # - Leader: flies its own mission or is manually controlled # - Follower: queries leader position + maintains offset # Configuration in config.csv: # Set 'swarm' column to group ID (1, 2, 3...) # Set 'is_leader' to 1 for the leader of each group # Followers maintain their (x, y, z) offset relative to leader # Multi-group example: # Group 1: drone 1 (leader), drones 2-4 (followers) # Group 2: drone 5 (leader), drones 6-8 (followers) # Groups operate independently — fly different missions # Communication: followers poll leader position via GCS server # Mesh radio recommended for field deployment (see Mesh Guide)Leader-Follower

# ═══ REAL HARDWARE CHECKLIST ═══ # Per-drone hardware: # □ PX4 flight controller (Pixhawk, Orqa H7, etc.) # □ GPS module (RTK preferred for ≤2cm accuracy) # □ Companion computer (RPi 4/5 is sufficient for MDS) # □ WiFi or mesh radio (all drones + GCS on same network) # □ LED strip (WS2812B) for light shows — controlled via companion # □ Unique HW_ID set on each companion (file or env var) # Per-drone software: # □ PX4 firmware (latest stable) # □ Python 3.11+ on companion # □ MDS installed: git clone + pip install -r requirements.txt # □ config.csv with correct HW_ID and formation offsets # □ shapes/ folder with choreography CSV # GCS: # □ Laptop running gcs-server (Flask + React dashboard) # □ Connected to same network as all drones # □ QGroundControl as backup (multi-vehicle view) # Critical PX4 parameters (per drone): # COM_RCL_EXCEPT = 4 (allow offboard without RC) # NAV_RCL_ACT = 0 (RC loss → continue mission) # MAV_SYS_ID = HW_ID (unique per drone)Hardware

# ═══ TIME SYNC — THE MAKE-OR-BREAK FOR SHOWS ═══ # Option 1: GPS PPS (best — sub-microsecond) # PX4 syncs internal clock to GPS time automatically # Companions sync via chrony or gpsd + PPS kernel module # Option 2: NTP over mesh network (good — 1-10ms) # GCS runs NTP server, all companions sync to it sudo apt install chrony # /etc/chrony/chrony.conf on GCS: # local stratum 1 # allow 192.168.168.0/24 # On each drone companion: # server 192.168.168.1 iburst # Option 3: MDS built-in sync (simplest) # MDS uses GCS server time as reference # Each drone fetches server time at startup and offsets # Good enough for formations. Not precise enough for tight shows. # Rule of thumb: # 10ms sync error = ~0.3m position error at 30 m/s # For shows with 2m separation: NTP is fine # For shows with 0.5m separation: need GPS PPSTime Sync

# ═══ 3D PREVIEW IN BROWSER ═══ git clone https://github.com/lucianotonet/3d-drone-show.git cd 3d-drone-show npm install npm run dev # Open http://localhost:5173 # Load your choreography CSV # Scrub timeline, rotate view, check spacing # Red dots = collision risk (drones closer than safety threshold) # Alternative: MDS GCS dashboard has built-in 2D formation view # Alternative: Blender viewport (if using SkyBrush)Preview

# ═══ INDOOR SWARM WITH OPTITRACK / VICON ═══ # Motion capture system provides position of each drone at 120+ Hz # Pipe to PX4 via MAVROS vision_pose (same as SLAM guide) # The Lumifly approach (sumietgore/DroneShow): # 1. ROS master on GCS PC # 2. Each drone runs MAVROS + position relay node # 3. Choreography node publishes target positions per drone # 4. Position controller on each drone tracks the target # Without motion capture — use LiDAR SLAM (see SLAM Guide): # Each drone runs FAST-LIO2 → MAVROS → PX4 LOITER mode # Swarm coordinator sends target positions over mesh/WiFi # Drone-to-drone relative positioning via UWB beacons (DWM1001) # UWB ranging for relative formation: # Each drone has a DWM1001 module ($15) # Measures distance to neighbors (±10cm accuracy) # No infrastructure needed — peer-to-peer rangingIndoor