Electrolyte imbalances are critical factors in major organ failure, contributing to an average of 13 deaths daily in the United States. Thus, developing an accurate, real-time in-body biochemical sensing system is vital for effective disease management and early organ failure detection. Current methodologies' limitations are: 1) blood draws are time-consuming, one-time measurements; 2) wearable biosensors have detection delays with limited sensing capabilities; and 3) current implantables are bulky, battery-powered, with limited selectivity. We are developing a battery-free, implantable biochemical sensing system designed for localized organ monitoring. The system includes a flexible wearable paired with a passive implantable that uses inductive coupling for wireless power transfer and data transmission. We were previously challenged with low sensitivity, poor sensor performance, and vulnerability to distance-induced distortions on the signal. Here, we introduce an optimized dual-pair system that accurately measures the sensor response by removing baseline variations caused by distance distortions. Our approach leverages a reference implantable to reject environmental variations, thereby enabling isolation of the sensor response. The implantable was redesigned to minimize sensor noise, achieving a 4x improvement in signal-to-noise ratio. Mathematically informed impedance matching enhanced the system's sensitivity by 340%, and the wearable was further miniaturized by 32%. Mathematical modeling paired with a distance rejection algorithm provided accurate vertical distance prediction and enabled the back-calculation of the analyte concentration from the sensor. This work validates a low-noise, high-sensitivity wireless sensing system that resolves a critical barrier by removing environmental variations from sensor readings, enabling accurate in-body concentration determination.

Woven shells combine the benefits of the robustness of domes and low shear resistance of typical weaving patterns, producing complex mechanical responses. Woven domes with radial and circumferential monofilaments show unique axial twisting coupled with compression, as well as exceptional frictional dissipation, leading to multistable states. Despite the advantages, woven shells currently lack an automated fabrication process which would increase precision, consistency, and design flexibility. We propose the use of dual extrusion Fused Deposition Modeling (FDM) 3D printing to create woven shells, which are difficult or infeasible to produce manually. We use Thermoplastic Polyurethane (TPU) as the base material, which allows repeated cycles of bending and stretching of thin shell features, and Polyvinyl alcohol (PVA) as the sacrificial material. We first perform indentation tests on 3D-printed standard woven shells to determine that the deformation and force-displacement responses are similar to their hand-woven counterparts. We then explore complex shell geometries with different shapes, variations in wire designs and orientations, and hybrid shells combining shells with different curvatures. We also use finite element analysis to validate the experimental results. The measured data shows that 3D printing of woven shells unlocks a wide design space producing varied force-displacement responses, including monotonically increasing, plateau, and snap-through behavior, and angles of twist responses, including clockwise, counter-clockwise and mixed rotations. Our 3D printed woven shells are promising energy-absorbing structures due to their tunability, lightweight design, repeatability, and high hysteresis, which makes it ideal for vehicular safety, protective sports gear, packaging, construction, soft robotics, and space applications.

Cooperative autonomous driving (CDA) relies on integrating real-world sensing with high-fidelity simulation to study perception, coordination, and decision-making in safe, repeatable conditions. Yet current infrastructure-assisted CDA research remains limited, particularly in bridging continuous multi-modal sensing with realistic, interactive simulation. This poster presents Live2Sim, a modular pipeline that streams live data from roadside units, connected autonomous vehicles (CAVs), and other sources into simulation in real time. Live2Sim standardizes heterogeneous sensor inputs and synchronizes simulated actors via ROS communication, faithfully mirroring complex traffic scenes for visualization, interaction, and scenario replay.    To extend beyond live experiments, Live2Sim was used to construct the V2X_UCLA dataset, featuring diverse traffic scenarios with varied intersection geometries, occlusions, dense flows, and long-tail events. Compared with existing V2X datasets, V2X_UCLA offers broader coverage, richer multi-modal sensing from both vehicles and infrastructure, and consistent temporal tracking. It supports not only cooperative perception but also downstream tasks such as trajectory prediction, motion planning in mixed traffic, and large-scale CDA strategy evaluation under diverse conditions.    The pipeline and dataset were validated by integrating Live2Sim into the CDA-SimBoost framework and replaying V2X_UCLA scenarios for perception and planning. Real-world bounding boxes were overlaid onto simulated images and point clouds, confirming close spatial and temporal alignment. Multi-sensor consistency was also tested by cross-validating simulated outputs across cameras, LiDAR, and other sensors. Together, these visualizations and real-time synchronization demonstrate that Live2Sim and V2X_UCLA provide a robust foundation for reproducible CDA research.  

  Metabolic fluxes are turnover rates of biochemical reactions in living organisms. Flux quantitation provides detailed insight into biological state and function. While a detailed protocol exists for in vitro flux quantitation, calculating in vivo fluxes poses additional challenges in sampling at designated time points without disrupting the biological state of the organism. The proposed method includes the use of five uniquely labeled glucose tracers infused at defined time points, which allowed for metabolite quantification from tissue samples in each mouse, a mammalian model system. The resulting labeling distribution allows for flux quantification using analytical mass balances from upper glycolysis. Research this summer included developing and testing a new Liquid Chromatography-Mass Spectrometry (LC-MS) method to separate hexose phosphate isomers, and refining a mathematical model of glucose mixing in MatLab. Separation between hexose phosphate isomers enables distinction between glucose 6-phosphate and glucose 1-phosphate, metabolites that are part of the glycogen synthesis pathway. Distinguishing between these compounds enables quantification of glycogenic fluxes, and a greater understanding of glucose utilization across tissues. The proposed method demonstrated success in separating hexose phosphate isomers in tissue samples from in vivo experiments. Incorporating a mixing coefficient into the mathematical model of glucose mixing accounts for non-instantaneous tracer mixing observed experimentally. Future research on the effects of LC-MS solvent pH must be conducted to refine separation of glucose 1-phosphate and fructose 6-phosphate. This study demonstrates the exciting potential for flux quantitation in clinical settings. Furthermore, the implications of a more nuanced understanding of glycogenic fluxes are far-reaching, and have the potential to improve understanding of the progression of metabolic diseases such as diabetes.  

In earlier radiofrequency communications, different transmitters used different frequency bands, and devices could listen to a specific transmitter by listening to that transmitter’s designated frequency band. However, with devices like drone controllers, relying on this technique creates a vulnerability to frequency jamming. Thus, drone controllers often hop between different frequencies to avoid being jammed on a specific frequency. With frequency hopping, the method of listening to a designated frequency no longer works. New methods such as WHIRLS solve this problem by taking a radio spectrum scan and identifying the time and frequency ranges of each signal from the raw I/Q data. These methods are accurate but slow, often taking several seconds to process milliseconds of data. In this paper, we propose a spectrogram-based floodfill algorithm to detect the time and frequency ranges of transmitters in real-time. We evaluated this algorithm’s average runtime compared to WHIRLS on the UAVSig drone dataset and compared each method’s accuracy through the number of signals detected, total signal area detected and false detection rate. We show that the floodfill algorithm far outclasses WHIRLS in terms of speed, decreasing runtime by 96%, for a minimal sacrifice in accuracy. Furthermore, the floodfill algorithm maintains its accuracy in low signal-to-noise ratio 8/15/2025 Johnny Zhang (SNR) situations, and in situations where multiple transmitters have different SNRs. Due to its high performance, the floodfill algorithm could prove effective in real-time signal analyzers for drones, Bluetooth devices and telecommunications.