skip to main content

Search for: All records

Creators/Authors contains: "Huang, G."

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Adults aged 65 years and older are the fastest growing age group worldwide. Future autonomous vehicles may help to support the mobility of older individuals; however, these cars will not be widely available for several decades and current semi-autonomous vehicles often require manual takeover in unusual driving conditions. In these situations, the vehicle issues a takeover request in any uni-, bi- or trimodal combination of visual, auditory, or tactile alerts to signify the need for manual intervention. However, to date, it is not clear whether age-related differences exist in the perceived ease of detecting these alerts. Also, the extent tomore »which engagement in non-driving-related tasks affects this perception in younger and older drivers is not known. Therefore, the goal of this study was to examine the effects of age on the ease of perceiving takeover requests in different sensory channels and on attention allocation during conditional driving automation. Twenty-four younger and 24 older adults drove a simulated SAE Level 3 vehicle under three conditions: baseline, while performing a non-driving-related task, and while engaged in a driving-related task, and were asked to rate the ease of detecting uni-, bi- or trimodal combinations of visual, auditory, or tactile signals. Both age groups found the trimodal alert to be the easiest to detect. Also, older adults focused more on the road than the secondary task compared to younger drivers. Findings may inform the development of next-generation of autonomous vehicle systems to be safe for a wide range of age groups.« less
  2. The availability of high quality surface observations of precipitation and volume observations by polarimetric operational radars make it possible to constrain, evaluate, and validate numerical models with a wide variety of microphysical schemes. In this article, a novel particle-based Monte-Carlo microphysical model (called McSnow) is used to simulate the outer rain bands of Hurricane Dorian which traversed the densely instrumented precipitation research facility operated by NASA at Wallops Island, Virginia. The rain bands showed steady stratiform vertical profiles with radar signature of dendritic growth layers near −15 °C and peak reflectivity in the bright band of 55 dBZ along withmore »polarimetric signatures of wet snow with sizes inferred to exceed 15 mm. A 2D-video disdrometer measured frequent occurrences of large drops >5 mm and combined with an optical array probe the drop size distribution was well-documented in spite of uncertainty for drops <0.5 mm due to high wind gusts and turbulence. The 1D McSnow control run and four numerical experiments were conducted and compared with observations. One of the main findings is that even at the moderate rain rate of 10 mm/h collisional breakup is essential for the shape of the drop size distribution« less
  3. Evaluating generative adversarial networks (GANs) is inherently challenging. In this paper, we revisit several representative sample-based evaluation metrics for GANs, and address the problem of how to evaluate the evaluation metrics. We start with a few necessary conditions for metrics to produce meaningful scores, such as distinguishing real from generated samples, identifying mode dropping and mode collapsing, and detecting overfitting. With a series of carefully designed experiments, we comprehensively investigate existing sample-based metrics and identify their strengths and limitations in practical settings. Based on these results, we observe that kernel Maximum Mean Discrepancy (MMD) and the 1-Nearest- Neighbor (1-NN) two-samplemore »test seem to satisfy most of the desirable properties, provided that the distances between samples are computed in a suitable feature space. Our experiments also unveil interesting properties about the behavior of several popular GAN models, such as whether they are memorizing training samples, and how far they are from learning the target distribution.« less