Autonomous Minefield Navigation Swarm

The Challenge

The Robofest Minefield Navigation Challenge required a swarm of micro aerial vehicles under 500g to safely guide a person at risk across a 100-meter field of buried explosives without external data links. Initial testing with commercial flight controllers such as Pixhawk and Diatone Mamba failed because their internal failsafes prevented arming in GPS-denied environments and rejected external visual odometry. We had to build a custom flight architecture from scratch that could handle complex A* pathfinding and thermal detection while remaining strictly under the weight limit.

Technical Deep Dive

Hardware Architecture

• 3D Printed Chassis: Used PLA for a Quad-X frame to minimize vibration and maximize the thrust-to-weight ratio.
• Custom Payloads: Designed 3D-printed, vibration-isolated mounts for the MLX90640 thermal arrays and a custom liquid dispenser mechanism using an SG90 servo for the marker drone.
• Avionics Deck: Engineered a custom PCB carrier board for the Teensy 4.0/4.1, integrating I2C/SPI interfaces for the MPU-6050, MTF-01 optical flow, and DPS310 barometer.

Firmware & Control

• Custom Teensy FC: Developed a proprietary C++ flight controller using a dual-core architecture to run PID loops at 400 Hz without telemetry interruption.
• Sensor Fusion: Implemented a complementary filter to fuse IMU and optical flow data, enabling rock-solid loiter capabilities indoors without GPS.
• Swarm Intelligence: Deployed a Raspberry Pi 4 master node running Python and OpenCV to recognize military hand gestures such as START, HOLD, and MOVE and dispatch UART velocity targets to the slave drones.

Results

• Competition Success: The system successfully secured a finalist position and a ₹2 lakh prize at Robofest Gujarat 5.0.
• Weight Optimization: Successfully integrated complex computing and payload systems while keeping the master drone at about 469g and scout drones at about 397g, well under the 500g limit.
• Thermal Pathfinding: Validated an A* algorithm that dynamically generates a low-risk corridor by calculating a probabilistic cost map based on thermal inertia footprints of buried objects.