Saturday, November 21, 2026

 

Robotic search and rescue in earthquake-collapsed buildings: capabilities, evidence gaps, and a roadmap for human-multi-robot rescue




Human–robot collaboration in earthquake rescue: seven classes of robots support human teams with aerial reconnaissance, rubble inspection, survivor searches, debris clearance and casualty transport.

image: 

Human–robot collaboration in earthquake rescue: seven classes of robots support human teams with aerial reconnaissance, rubble inspection, survivor searches, debris clearance and casualty transport.

view more 

Credit: Defeng Xu, Feifei Sun∗, Songhang He, Xiangdian Chen, Yuqing Gao∗, Zhen Wang / Tongji University, China; Haiying Wang / Ministry of Emergency Management, China






Robots can explore dangerous earthquake rubble and detect possible signs of trapped survivors, but turning these findings into safe rescue operations remains a major challenge. In a review published in Smart Construction, researchers from Tongji University and collaborating institutions assessed seven categories of rescue robots and identified critical gaps between search and rescue. Drawing on evidence from 144 core sources, the study proposes a human-supervised approach in which different robots share information and coordinate their tasks. The findings highlight priorities for developing safer and more effective robotic rescue systems.

After an earthquake, rescue teams often face unstable debris, narrow gaps and limited visibility as they search for people trapped inside collapsed buildings. Robots can enter some of these spaces and collect information without exposing rescuers to the same risks. Yet finding signs of a survivor is not the same as reaching that person safely. A review published on 16 November 2026 in Smart Construction examines why this gap persists. Conducted by researchers at Tongji University, Jilin Agricultural University and China's National Earthquake Response Support Service, the study brings together evidence on what rescue robots can currently achieve and what remains largely untested.

Drawing on 144 core sources, the researchers assessed seven categories of robots: drones; wheeled, tracked and reconfigurable machines; quadruped robots; snake-like and continuum robots; humanoid and hybrid robots; heavy-duty engineering machines; and casualty extraction and medical transport systems. Their central conclusion is that progress has been uneven. There is substantial field evidence for reconnaissance, mapping and preliminary searches, including detecting possible signs of life. Much less is known about how reliably robots can maintain communications, deliver supplies, assist with debris removal near trapped people or move casualties to safety.

For rescuers, this difference has practical consequences. A camera, thermal sensor or radar may indicate that a person is present beneath rubble. But a possible signal does not necessarily reveal the precise three-dimensional position needed to guide excavation. Without reliable positioning and an assessment of structural hazards, an intervention could place the survivor at further risk.

To connect search and rescue, the team organized the process into seven stages: entering the rubble, moving through it, mapping the scene, detecting signs of survivors, confirming their locations, providing communications and logistical support, and carrying out rescue interventions and casualty care. The stages may overlap as conditions change. The framework envisages robots working in sequence or alongside one another according to their strengths. A drone might survey damaged buildings and identify possible access routes. A tracked robot could inspect debris at ground level, while a slender snake-like robot could enter cavities that larger machines cannot reach. Human rescue teams would then decide how to proceed.

Cooperation, however, requires more than deploying several machines at the same site. Robots may use different maps, coordinate systems and communication links. Information gathered by one platform must remain meaningful to the next, especially when visibility is poor, signals are interrupted or the rubble shifts. The researchers therefore propose that each task handover include not only images and locations but also their sources, timestamps and uncertainties. Records of structural hazards, access routes, communication conditions and current mission status should also travel with the data. Crucially, actions that could affect survivor safety would require human authorization.

The review also draws attention to the difference between experimental success and readiness for disaster response. To assess this, the team used two separate scales: one describes the level of validation, from simulations to real-world deployment; the other indicates how closely those tests resemble searches inside earthquake rubble. A deployment after the 2017 Mexico City earthquake illustrates the distinction. A snake-like robot entered a collapsed building and sent video from two interior cavities. This demonstrated that the machine could inspect otherwise difficult-to-access spaces, but it did not find anyone and provided no evidence of a completed rescue.

Similar limits apply to other robot types. Quadruped robots have navigated challenging terrain in trials, and remote-controlled engineering machines can move heavy materials. Neither achievement, on its own, proves that the equipment can operate safely beside a trapped person in unstable rubble. The final stages remain particularly demanding. Removing debris can alter the stability of a damaged structure, while transporting an injured person requires careful control of pressure, body support and vibration. Research on casualty-handling robots has largely relied on prototypes, mannequins and controlled tests rather than evacuations during real earthquake disasters.

These findings place trained responders at the centre of future rescue systems. Robots could help gather evidence, explore inaccessible spaces and support specific operations, while decisions involving structural safety, patient contact and other high-risk interventions remain under human supervision. The goal is to strengthen rescue teams, not replace them. Better coordination could also prove useful outside immediate emergency response. Shared mapping and reliable data exchange may support inspections of damaged buildings and help engineers plan access routes, monitor hazards and organize repairs. The researchers suggest that similar principles could benefit other hazardous construction and recovery tasks.

The authors emphasize that their work is a review of existing evidence, not a demonstration of a fully functioning robot rescue team. Their proposed workflow has yet to be tested from beginning to end in a real earthquake. Future trials should measure positioning accuracy, information lost during handovers, communication failures and forces applied to injured people. By focusing on the links between detection, decision-making and intervention, the study shifts attention from what an individual robot can do to what a coordinated team can achieve. For earthquake rescue, the next advance may depend as much on reliable information and human oversight as on new machines.

This paper was published in Smart Construction.

Xu DS, un F, He S, Chen X, Gao Y, et al. Robotic search and rescue in earthquake-collapsed buildings: capabilities, evidence gaps, and a roadmap for human-multi-robot rescue. Smart Constr. 2026(4):0020, https://doi.org/10.55092/sc20260020.

DOI: 10.55092/sc20260020

No comments: