Abstract

The purpose of this study is to design a fire detection drone system with a unique hardware-in-the-loop (HIL) simulation architecture, mainly focusing on the search and localization algorithms and simulating thermal cameras to test computer vision-based detection algorithms. The autopilot hardware has been designed exclusively for this research work. The basic flight algorithm has been implemented in the autopilot firmware. To communicate and configure the autopilot, a ground control station (GCS) is developed. The GCS exchanges data with autopilot hardware using a serial port for both telemetry and HIL data links. A game engine (Unity3D) is used for implementing the simulator’s 3D graphics. To solve the rigid-body equations, the Unity3D built-in Nvidia PhysX system is utilized. The simulator exchanges data with the GCS using a UDP port. The GCS acts as a bridge between autopilot and simulator. To achieve real-time simulation performance, in most of the simulation systems and the GCS, multitasking is implemented. Furthermore, a simulated thermal camera with a raw image provider (similar to the actual hardware output) and a fire-making system in a forest-like environment has been developed to set fire to the simulated forest either at a specific location or randomly. The system consistency has been tested by performing some simulation tests and furthermore by testing the system in a real flying platform and testing the drone outdoor. Finally, the outcome of the system exhibited a good agreement with the autopilot as well as the guidance and navigation system in terms of the fire detection and positioning algorithms.