Vision-Based Control of UAV for Autonomous Firefighting

Urban and industrial environments frequently necessitate the immediate readiness of firefighting personnel to address potential fire emergencies, which can lead to a perpetual shortage of available manpower. To counter this challenge, we propose an integrated approach to autonomous firefighting through the utilization of Unmanned Aerial Vehicles (UAVs). The UAVs serve a dual purpose: providing auxiliary support to conventional firefighting efforts while simultaneously mitigating risks to human life. Our proposed methodology incorporates deep learning-based fire detection. The system combines image feature analysis with data obtained from distance sensor to establish Cartesian coordinates of the identified fire sources. Autonomous vision-based control for multirotor platform is developed, guided by the extracted image features and Cartesian coordinates. This system is designed to facilitate both swift deployment and autonomous operation, while still allowing for manual intervention as necessary. We present experimental validations for the vision-based control of the multirotor platform under outdoor conditions. Further tests are conducted to assess the performance characteristics of the spray subassembly, including spray distance and flow rate, prior to its deployment on aerial platforms. The experimental results are discussed for performance analysis of the approach.