Design of Autonomous Rover for Firefighter Rescue: Integrating Deep Learning with ROS2

Abstract

Firefighters often face the dangerous task of navigating burning structures to rescue endangered individuals, with challenges like intense heat, toxic fumes, and debris. We introduce a rover concept to aid these rescue missions through autonomous data collection. This rover can climb stairs, detect distressed sounds, locate individuals, and gather environmental data. The paper details the design of its power and hardware components for data operations and movement. Using the Robotic Operating System (ROS2), telemetry data is automated in real-time, integrating three NVIDIA Jetson Nanos and two deep learning models. These models identify individuals via infrared (IR) video and detect human screams. We discuss the models’ performance, latency, and integration with the ROS2 framework. We aim to expedite rescue operations, enhancing safety and survival rates for firefighters and civilians.