After an earthquake, wildfire, or other major disaster, first responders may enter locations where smoke, debris, or buildings obscure their surroundings, cell towers are damaged, and the Global Positioning System (GPS) is unreliable. Restoring communications is critical, but connectivity alone cannot tell rescuers where people are or how conditions are changing.

UC Santa Barbara electrical and computer engineering professor Upamanyu Madhow is leading research aimed at helping wireless networks fill that information gap, using the same signals that connect people and devices to locate individuals and detect movement, even when visibility is limited. 

Madhow is a co-principal AReS, short for Assessing, Restoring, and Supporting Post-Disaster Connectivity and Response, a new $4.1 million National Science Foundation project led by Princeton University. Funded through NSF’s Verticals-enabling Intelligent Network Systems (VINES) program, AReS combines artificial intelligence, augmented reality, and integrated sensing and communication, or ISAC, an approach that uses wireless networks not only to transmit data, but also to sense their surroundings. 

Madhow leads the project’s work on network-assisted positioning and tracking. His group will work closely with researchers at Princeton, Duke, and Rice to develop algorithms for locating devices, detecting people who are not carrying transmitting devices, and tracking movement, and to prototype and demonstrate those approaches on experimental wireless systems. 

“Wireless networks are already deployed nearly everywhere,” Madhow said. “That gives us an opportunity to use cellular systems for distributed sensing as well as communication.”

Building on his group’s expertise in signal processing, location estimation, and precise synchronization of wireless equipment, Madhow is addressing two challenges: finding someone who can transmit a wireless signal and detecting people who have no functioning device at all. 

Finding Someone from their Emitted Signal
A first responder carrying wireless equipment or a disaster victim whose phone can still transmit may be able send a signal even when GPS cannot establish their location. Madhow’s team is investigating how the network itself could determine where the signal originated. 

Multiple receives would listen for the same transmission, each gathering clues about when the signal arrived, the direction it came from, and the path it traveled. These measurements would combine those measurements to estimate the device’s location. 

Buildings and other obstacles complicate the process because wireless signals can reach receivers by several routes, bouncing off surfaces, referred to as reflecting, rather than traveling in straight lines. By combining the timing and direction of those signals with a three-dimensional map of the area, useful information can be extracted from potentially confusing reflections. 

The measurements will not always point to a single answer, especially in a damaged or complicated environment. Madhow’s team will use AI and statistical methods to weigh the available clues and determine which locations are most likely, even when some information is incomplete or uncertain. 

The approach is conceptually similar to earlier research by Madhow’s group, which used 3D maps to interpret distorted satellite signals and improve GPS positioning in cities. That technology was commercialized through the startup ShadowMaps, which Uber acquired in 2016. 

Although the signals and technical methods in AReS differ from those used in ShadowMaps, it also draws on knowledge of the surrounding environment to improve location estimates. AReS also takes full advantage of the new possibilities opened up by advances in wireless technology and computing over the past decade. Modern cellular networks increasingly use multiple antennas and greater bandwidth, providing more detailed information about how signals travel through an environment. At the same time, advances in computing and AI allow researchers to process those measurements and model complicated signal paths with greater sophistication using “digital twins” of the environment.

Using a technique known as ray tracing, researchers can simulate how wireless signals travel and reflect off buildings and other surfaces. Machine-learning models could help interpret the resulting patterns and identify a device’s likely location, even when signals have traveled along indirect paths. 

Finding Someone Without a Device
Locating someone trapped beneath rubble or inside a damaged building presents another challenge when that person has no working phone or other means of sending a signal. 

For these situations, the researchers will explore device-free multi-static sensing. Rather than listening for a signal from the person, one wireless node would transmit signals into an area while receivers at other locations measure how those signals interact with people and objects. Comparing measurements from several viewpoints could allow the system to detect and track movement without requiring the person to carry a functioning device. 

Buildings, furniture, rubble, and other stationary objects create significant interference, so AReS researchers will initially focus on detecting motion. 

“At a coarse level, it can be used to determine where there are people who need help,” Madhow said. “At a fine-grained level, sensors placed close enough to the scene could potentially detect small amounts of motion or even breathing, behind walls or rubble.”
The project will also explore whether relatively simple receivers placed around a scene can work with more powerful cellular equipment, potentially making this type of distributed sensing more practical to deploy.

From Research to Disaster Response
Turning these techniques into tools that first responders can use will require more than advances in positioning and sensing. AReS researchers must also find ways to restore damaged communication networks, combine data from multiple systems, and present information in ways that are useful during an emergency.

Madhow, Princeton professor Yasaman Ghasempour, and Duke professor Tingjun Chen were already collaborating on fundamental wireless-systems research when the call for proposals came from NSF. The larger AReS team brings together expertise in wireless networks, artificial intelligence, sensing, augmented reality, and system integration.  

Over the next three years, the researchers aim to integrate their technologies and demonstrate a system capable of restoring connectivity while collecting and sharing information about a disaster area. 

“Our goal is to make the cellular network useful not only for restoring communication, but also for helping first responders understand what is happening around them in real time,” Madhow said. “If we can help responders locate people and better understand conditions in places where GPS, visibility, and infrastructure are compromised, these technologies could become an impactful tool for disaster response.”