Abstract
In the wake of catastrophic earthquakes, rescue operations encounter significant obstacles in locating and reaching individuals trapped under debris. This study introduces the Smart Earthquake Rescue Robot (SERR) prototype, a cutting-edge solution designed to enhance the efficiency and effectiveness of earthquake rescue missions. The SERR is a mobile robot with advanced features, including live video streaming through an integrated camera, Grid-Eye temperature detection, and provisions for communication via built-in speakers and microphones, although these audio communication capabilities are pending implementation in the current prototype. It can be remotely controlled using a smartphone application, offering a safer and more efficient method for conducting rescue operations. Unlike recent advances discussed in the literature, SERR uniquely combines visual, thermal, and audio data with a multi-modal Convolutional Neural Network with Long Short-Term Memory (CNN-LSTM) model (RescueNet), achieving high accuracy (0.94), precision (0.90), recall (0.96), and F1-score (0.92) in detecting survivors, as validated by MATLAB simulations using USGS-PAGER data. The SERR's rapid runtime (35 ms) highlights its promise as a tool to improve earthquake rescue outcomes.