Efficient simulation is essential for enhancing proactive preparedness for sudden-onset disasters such as earthquakes.Recent advancements in large language models (LLMs) as world models show promise in simulating complex scenarios.This study examines multiple LLMs to proactively estimate perceived earthquake impacts.Leveraging multimodal datasets including geospatial, socioeconomic, building, and street-level imagery data, our framework generates Modified Mercalli Intensity (MMI) predictions at zip code and county scales.Evaluations on the 2014 Napa and 2019 Ridgecrest earthquakes using USGS "Did You Feel It? (DYFI)" reports demonstrate significant alignment, as evidenced by a high correlation of 0.88 and a low RMSE of 0.77 as compared to real reports at the zip code level.Techniques such as retrieval-augmented generation (RAG) and in-context learning (ICL) can improve simulation performance, while visual inputs notably enhance accuracy compared to structured numerical data alone.These findings show the promise of LLMs in simulating disaster impacts that can help strengthen preevent planning.

Distributed Acoustic Sensing (DAS) that transforms city-wide fiber-optic cables into a large-scale strain sensing array has shown the potential to revolutionize urban traffic monitoring by providing a fine-grained, scalable, and low-maintenance monitoring solution. However, the real-world application of DAS is hindered by challenges such as noise contamination and interference among closely traveling cars. In response, we introduce a self-supervised U-Net model that can suppress background noise and compress car-induced DAS signals into high-resolution pulses through spatial deconvolution. Our work extends recent research by introducing three key advancements. Firstly, we perform a comprehensive resolution analysis of DAS-recorded traffic signals, laying a theoretical foundation for our approach. Secondly, we incorporate space-domain vehicle wavelets into our U-Net model, enabling consistent high-resolution outputs regardless of vehicle speed variations. Finally, we employ L-2 norm regularization in the loss function, enhancing our model’s sensitivity to weaker signals from vehicles in remote traffic lanes. We evaluate the effectiveness and robustness of our method through field recordings under different traffic conditions and various driving speeds. Our results show that our method can enhance the spatial-temporal resolution and better resolve closely traveling cars. The spatial deconvolution U-Net model also enables the characterization of large-size vehicles to identify axle numbers and estimate the vehicle length. Monitoring large-size vehicles also benefits imaging deep earth by leveraging the surface waves induced by the dynamic vehicle-road interaction.

In-home gait analysis is important for providing early diagnosis and adaptive treatments for individuals with gait disorders. Existing systems include wearables and pressure mats, but they have limited scalability due to dense deployment and device carrying/charging requirements. Recently, vision-based systems have been developed to enable scalable, accurate in-home gait analysis, but it faces privacy concerns due to the exposure of people's appearances and daily activities. To overcome these limitations, our prior work developed footstep-induced structural vibration sensing for in-home gait monitoring, which is device-free, wide-ranged, and perceived as more privacy-friendly. Although it has succeeded in temporal parameter estimation, it shows limited performance for spatial gait parameter estimation due to the low accuracy in footstep localization. In particular, the localization error mainly comes from the estimation error of the wave arrival time at the vibration sensors and its error propagation to wave velocity estimations. To this end, we present GaitVibe+, a vibration-based footstep localization method fused with temporarily installed cameras for in-home gait analysis. Our method has two stages: fusion and operating stages. In the fusion stage, both cameras and vibration sensors are installed to record only a few trials of the subject's footstep data, through which we characterize the uncertainty in wave arrival time and model the wave velocity profiles for the given structure. In the operating stage, we remove the camera to preserve privacy at home. The footstep localization is conducted by estimating the time difference of arrival (TDoA) over multiple vibration sensors, whose accuracy is improved through the reduced uncertainty and velocity modeling during the fusion stage. We evaluate GaitVibe+ through a real-world experiment with 50 walking trials. With only 3 trials of multi-modal fusion, our approach has an average localization error of 0.22 meters, which reduces the spatial gait parameter error by 4.1x (from 111.4% to 27.1%) compared to the existing work.

Non-intrusive load monitoring (NILM) is the set of algorithmic techniques for inferring the operational states of individual appliances in a household given the aggregate electrical measurements at a single point of instrumentation. Most successful techniques to-date approach the problem from a supervised learning perspective and thus rely on labeled data, which is costly to obtain, and assume similar data distributions for appliances beyond those in the training set. To alleviate this problem, we formulated NILM in a domain adaptation context. Using Binary OnLine FactorizaTion (BOLT) as the baseline model, we first demonstrate that direct application of domain adversarial training without application-specific modifications is unsuccessful, and hypothesize that this may be due to differences in the data distributions. We then propose Dyna-BOLT: a domain-adaptive variant of BOLT, in which we 1) provide private decoders for the source and target domains to account for the differences in current waveforms, and 2) tie the weights between the two decoders using a metric that was specifically trained to distinguish between appliance classes. We evaluate Dyna-BOLT on a publicly available dataset (REDD) and demonstrate that it compares favorably to unsupervised methods.

We introduce a physics-guided signal processing approach to extract a damage-sensitive and domain-invariant (DS & DI) feature from acceleration response data of a vehicle traveling over a bridge to assess bridge health. Motivated by indirect sensing methods' benefits, such as low-cost and low-maintenance, vehicle-vibration-based bridge health monitoring has been studied to efficiently monitor bridges in real-time. Yet applying this approach is challenging because 1) physics-based features extracted manually are generally not damage-sensitive, and 2) features from machine learning techniques are often not applicable to different bridges. Thus, we formulate a vehicle bridge interaction system model and find a physics-guided DS & DI feature, which can be extracted using the synchrosqueezed wavelet transform representing non-stationary signals as intrinsic-mode-type components. We validate the effectiveness of the proposed feature with simulated experiments. Compared to conventional time-and frequency-domain features, our feature provides the best damage quantification and localization results across different bridges in five of six experiments.

The objective of my thesis work is to develop physics-guided data-driven approaches for drive-by bridge health monitoring (BHM) that are scalable and eliminate the need of acquiring training data from every bridge. BHM allows us to detect bridge damage in earlier stages, which is essential for preventing more severe damage and collapses that may lead to significant business and human losses. Using vibrations from drive-by vehicles for BHM has various advantages, such as: economical and no need for on-site maintenance of equipment on bridges. However, many such approaches face analysis challenges for monitoring multiple bridges because 1) they either require labeled data from each bridge, which is expensive and time-consuming to collect, or 2) if we directly apply the supervised model trained for one bridge to other bridges, damage diagnostic accuracy could significantly reduce because of distribution mismatch between different bridges' data. In this work, I overcome the first challenge by leveraging physical insights about the vehicle-bridge interaction (VBI) to guide the data-driven approach and diagnose damage in a semi-supervised way. Also, I overcome the second challenge through a domain adversarial training and multi-task learning framework. The framework extracts features that are sensitive to damages and invariant across bridges to monitor multiple bridges without the need for training data from every bridge.