Integrating AI and Multimodal Sensor Fusion with Swarm Robotics for Survivor Detection in Collapsed Structures

This research investigates how artificial intelligence and swarm robotics can be combined with multimodal sensor fusion to enhance survivor detection in disaster environments, particularly collapsed urban structures. Recognising the limitations of traditional Urban Search and Rescue (USAR) methods—such as manual searches, canine units, and isolated technical devices—the paper presents a systematic exploration of how autonomous robotic swarms can be deployed to navigate, communicate, and localise survivors in complex, debris-filled environments. Through a review of existing technologies, it highlights the importance of decentralised control, robust sensing, and adaptive decision-making to improve the speed and accuracy of life detection after disasters such as earthquakes.

The paper examines a range of sensing technologies essential to USAR, including Doppler radar, thermal infrared, CO₂ and VOC sensors, acoustic arrays, seismic detectors, ultrasonic sensors, and LiDAR. By detailing the functionality and range of each sensor, the author demonstrates how their complementary strengths can be integrated through sensor fusion—where multiple data sources combine to create a unified and more reliable picture of survivor presence and location. Building on this, the study explores the behavioural mechanisms of swarm robotics, such as spatial organisation, collective navigation, and consensus-based decision-making, which enable a group of autonomous robots to coordinate without central oversight.

A key contribution of the research lies in its proposed framework for integrating AI-driven control with multimodal sensor data across a swarm network. By using machine learning models like deep reinforcement learning and SARSA, the framework allows each robot to learn optimal behaviours dynamically and cooperate with others for efficient coverage and detection. This adaptive intelligence, paired with fault-tolerant and scalable swarm design, positions the system as a transformative tool for next-generation disaster response. The paper concludes that AI-enabled swarm robotics, when supported by sensor fusion and real-time data processing, can significantly increase the probability of locating survivors quickly and safely—reducing dependence on human rescuers in hazardous environments and advancing the frontier of autonomous humanitarian technology.