Search for: All records

Abstract Differentiable rendering of translucent objects with respect to their shapes has been a long‐standing problem. State‐of‐the‐art methods require detecting object silhouettes or specifying change rates inside translucent objects—both of which can be expensive for translucent objects with complex shapes. In this paper, we address this problem for translucent objects with no refractive or reflective boundaries. By reparameterizing interior components of differential path integrals, our new formulation does not require change rates to be specified in the interior of objects. Further, we introduce new Monte Carlo estimators based on this formulation that do not require explicit detection of object silhouettes. 

This paper details the development and analysis of a computational neuroscience model, known as a Synthetic Nervous System, for the control of a simulated worm robot. Using a Synthetic Nervous System controller allows for adaptability of the network with minimal changes to the system. The worm robot kinematics are inspired by earthworm peristalsis which relies on the hydrostatic properties of the worm’s body to produce soft-bodied locomotion. In this paper the hydrostatic worm body is approximated as a chain of two dimensional rhombus shaped segments. Each segment has rigid side lengths, joints at the vertices, and a linear actuator to control the segment geometry. The control network is composed of non-spiking neuron and synapse models. It utilizes central pattern generators, coupled via interneurons and sensory feedback, to coordinate segment contractions and produce a peristaltic waveform that propagates down the body of the robot. A direct perturbation Floquet multiplier analysis was performed to analyze the stability of the peristaltic wave’s limit cycle. 

The recent report by American Society of Civil Engineers gave the nation's bridges an unimpressive C grade. Across the country, more than 617,000 highway bridges: 46,154 structurally deficient and 42% 50+ years old. Continuous bridge assessment is essential to protect public safety. Federal Highway Administration requires all highway bridges inspected once every 24 months. However, any drastic change on bridges within 24 months will be left undetected. Nonetheless, bridge inspection is time-consuming and labor-intensive. Civil engineers have been using bridge health monitoring (BHM) systems with wired and/or wireless sensors to measure structural response (e.g., displacement, strain, acceleration) of a bridge. The response measurements are then converted to the information related to structural health for assessment. State-of-the-art BHM technology deploys sensor networks to facilitate data connection. Installing cables is expensive and subject to extreme weather. Wireless solutions face challenges such as energy consumption. Sensors are battery-powered. Another not well-publicized problem is security threats inherited in wireless networks. Our approach to wireless BHM is to utilize sensors networkless by collecting data with a drone. Similar to a mail carrier who goes around and picks up the mail, a drone collects data from sensors throughout the bridge. A drone eliminates restrictions for civil engineers on node placement since the drone replaces sink nodes. Networkless makes BHM less prone to attacks such as Jamming and DoS. To secure access, we deploy a Needham-Schroeder authentication protocol for the drone to collect data from sensor nodes securely. Networkless sensing for BHM benefits energy efficiency. It saves battery life as the sensor nodes remain asleep until scheduled transmission or woken up by a drone. It reduces design complexity and operation energy. The system also assures security since there is no vulnerable network to be attacked. 

When subjected to the lap shear testing, spot welds created by brazing, resistance welding, or other techniques may fail either by a plug failure mode (also called pull-out mode) or an interfacial shear failure mode. In the past, plug failure mode was thought to be depend- ent on base metal ultimate tensile strength, spot diameter and plate thickness, while interfacial failure can be determined by interface shear strength and spot area. No fracture mechanics model or failure process is invoked in such an approach, and its predictive capability is often doubted compared to realistic experiments. This work conducts a parametric study to assess the failure behavior as a function of dominant three-dimensional geometric parameters based on the Gurson-Tvergaard-Needleman (GTN) damage mechanics model and no-damage mod- el respectively. Different necking conditions are considered as precursors to the two failure modes in the no-damage model. It is found out that a small ratio of spot diameter to plate thickness promotes interfacial shear failure while a large ratio favors plug failure. Other geometric parameters such as the filler interlayer thickness, if used, play a secondary role. The calculated peak force Fwt is not much different between the GTN and no-damage analyses, and better agreement is shown in the small nugget region. Normalized peak force calculated from the GTN model with the porosity f0 set to 0.01 showed the best agreement with pervious tensile shear tests on spot-welded DP980 lap joints in comparison to that calculated from the GTN model with f0 at 0.02 and the no-damage model. Note that heterogeneous distribution of materi- al strength across the joint region was considered in the GTN model, which was estimated based on the hardness map measured across the joint cross section. 

The prediction of Secondary Organic Aerosol (SOA) in regional scales is traditionally performed by using gas-particle partitioning models. In the presence of inorganic salted wet aerosols, aqueous reactions of semivolatile organic compounds can also significantly contribute to SOA formation. The UNIfied Partitioning-Aerosol phase Reaction (UNIPAR) model utilizes the explicit gas mechanism to better predict SOA formation from multiphase reactions of hydrocarbons. In this work, the UNIPAR model was incorporated with the Comprehensive Air Quality Model with Extensions (CAMx) to predict the ambient concentration of organic matter (OM) in urban atmospheres during the Korean-United States Air Quality (2016 KORUS-AQ) campaign. The SOA mass predicted with the CAMx-UNIPAR model changed with varying levels of humidity and emissions and in turn, has the potential to improve the accuracy of OM simulations. The CAMx-UNIPAR model significantly improved the simulation of SOA formation under the wet condition, which often occurred during the KORUS-AQ campaign, through the consideration of aqueous reactions of reactive organic species and gas-aqueous partitioning. The contribution of aromatic SOA to total OM was significant during the low-level transport/haze period (24-31 May 2016) because aromatic oxygenated products are hydrophilic and reactive in aqueous aerosols. The OM mass predicted with the CAMx-UNIPAR model was compared with that predicted with the CAMx model integrated with the conventional two product model (SOAP). Based on estimated statistical parameters to predict OM mass, the performance of CAMx-UNIPAR was noticeably better than the conventional CAMx model although both SOA models underestimated OM compared to observed values, possibly due to missing precursor hydrocarbons such as sesquiterpenes, alkanes, and intermediate VOCs. The CAMx-UNIPAR model simulation suggested that in the urban areas of South Korea, terpene and anthropogenic emissions significantly contribute to SOA formation while isoprene SOA minimally impacts SOA formation. 

Aliasing refers to the phenomenon that high frequency signals degenerate into com- pletely different ones after sampling. It arises as a problem in the context of deep learning as downsampling layers are widely adopted in deep architectures to reduce parameters and computation. The standard solution is to apply a low-pass filter (e.g., Gaussian blur) before downsampling [37]. However, it can be suboptimal to apply the same filter across the entire content, as the frequency of feature maps can vary across both spatial locations and feature channels. To tackle this, we propose an adaptive content-aware low-pass filtering layer, which predicts separate filter weights for each spatial location and chan- nel group of the input feature maps. We investigate the effectiveness and generalization of the proposed method across multiple tasks including ImageNet classification, COCO instance segmentation, and Cityscapes semantic segmentation. Qualitative and quanti- tative results demonstrate that our approach effectively adapts to the different feature frequencies to avoid aliasing while preserving useful information for recognition. Code is available at https://maureenzou.github.io/ddac/.