Using UWB Radios as Sensors for Disaster Recovery Jeongeun Julie Lee and Suresh Singh Department of Computer Science Portland State University Portland, OR 97207 Email: jeong.julie.lee@intel.com, singh@cs.pdx.edu Abstract—This paper considers the problem estimating the interior structure of a collapsed building by using embedded UWB radios as sensors. We created an extensive database of UWB propagation data through various building materials. Then, using this data and a novel algorithm, we demonstrate that we can indeed determine material type, thickness and cavity dimensions using UWB radios. This paper presents the algorithm and the evaluation results. As we show, for most common building materials such as concrete and reinforced concrete, the presented algorithm has a very small estimation error. I. I NTRODUCTION Rescuing trapped survivors after a building collapses fol- lowing an earthquake is a difficult and slow task. The primary reason for this is that rescuers do not know the structural stability of the rubble and thus need to excavate slowly and deliberately. For trapped survivors, this slow excavation can prove fatal. Tools to determine the location of survivors and the structural topology of the collapsed building typically consist of probes or robots lowered through holes. Unfortunately, these tools can only survey a piece of the whole structure and are typically restricted to mapping the top part of the collapse. Person trapped in loose rubble Rubble Wall-embedded sensor Sensor displaced from its wall mount Trapped survivors No sensors nodes here survived the collapse Fig. 1. Cutaway view of the building’s interior. In our research we are developing sensors that are embedded in the walls of the building during construction (or retrofitted later). These sensors consist of a UWB component for com- munication and sensing as well as an ultrasonic sensor and a b c d e f Reconstruction of wall based on signal attenuation and relative position information Sensor knows approximate damage to its retaining wall Fig. 2. Representation of the building’s interior produced by our system. heartbeat sensor. The idea is that when the building collapses, these sensors will collaboratively sense the interior of the collapsed structure and pass this information to the surface where software fuses the data to create a three dimensional view of the building’s interior. Figure 1 shows a view of the interior of the collapsed building. As shown, there are several survivors located in different cavities (at different depths) as well as a complicated arrangement of walls that form the support for the whole structure. It is easy to see that by carelessly removing the wrong piece of the rubble, we may precipitate a secondary collapse with fatal consequences. Thus, the goal of our system is to develop as detailed a map of the interior as possible in order to give rescuers hints about how to excavate. In a collapse, we expect many sensors to be destroyed. Furthermore, the location of the remaining sensors vis a vis the rubble, will limit the ability of our system to provide as detailed a map as illustrated in Figure 1. However, even a partial map such as Figure 2 that can be produced by our system will be very beneficial to rescuers. This map shows the location of various survivors as well as the primary support walls in the structure. It is easy to see that developing this form of a map requires us to solve several sensing problems including detection of people via heartbeats or co 2 levels, identification of cavity shapes, determination of the supporting