Search for: All records

Intelligent driving assistance can alert drivers to objects in their environment; however, such systems require a model of drivers' situational awareness (SA) (what aspects of the scene they are already aware of) to avoid unnecessary alerts. Moreover, collecting the data to train such an SA model is challenging: being an internal human cognitive state, driver SA is difficult to measure, and non-verbal signals such as eye gaze are some of the only outward manifestations of it. Traditional methods to obtain SA labels rely on probes that result in sparse, intermittent SA labels unsuitable for modeling a dense, temporally correlated process via machine learning. We propose a novel interactive labeling protocol that captures dense, continuous SA labels and use it to collect an object-level SA dataset in a VR driving simulator. Our dataset comprises 20 unique drivers' SA labels, driving data, and gaze (over 320 minutes of driving) which will be made public. Additionally, we train an SA model from this data, formulating the object-level driver SA prediction problem as a semantic segmentation problem. Our formulation allows all objects in a scene at a timestep to be processed simultaneously, leveraging global scene context and local gaze-object relationships together. Our experiments show that this formulation leads to improved performance over common sense baselines and prior art on the SA prediction task. 

Abstract During intense geomagnetic storms, the magnetopause can move in as far as geosynchronous orbit, leaving the satellites in that orbit out in the magnetosheath. Spacecraft operators turn to numerical models to predict the response of the magnetopause to solar wind conditions, but the predictions of the models are not always accurate. This study investigates four storms with a magnetopause crossing by at least one GOES satellite, using four magnetohydrodynamic models at NASA's Community Coordinated Modeling Center to simulate the events, and analyzes the results to investigate the reasons for errors in the predictions. Two main reasons can explain most of the erroneous predictions. First, the solar wind input to the simulations often contains features measured near the L1 point that did not eventually arrive at Earth; incorrect predictions during such periods are due to the solar wind input rather than to the models themselves. Second, while the models do well when the primary driver of magnetopause motion is a variation in the solar wind density, they tend to overpredict or underpredict the integrated Birkeland currents and their effects during times of strong negative interplanetary magnetic field (IMF)B_z, leading to poorer prediction capability. Coupling the MHD codes to a ring current model, when such a coupling is available, generally will improve the predictions but will not always entirely correct them. More work is needed to fully characterize the response of each code under strong southward IMF conditions as it relates to prediction of magnetopause location.