Browsing by Author "Rastogi, Aakanksha"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Implementation of a GIS with Analytics for Pollution Data
    (North Dakota State University, 2015) Rastogi, Aakanksha
    Particulate matter called pm2.5 refers to harmful microscopic liquid or solid matter of diameter 2.5 microns suspended in the air. To effectively analyze and visualize data for pollution category pm2.5, a Geographic Information System (GIS) was created. This application, named the Pollution Analysis System (PAS), was designed and implemented to help the user visualize the data in terms of graphs, charts and maps. The interactive visualizations are aimed to enhance data readability and give meaning to the raw data. Correlation techniques were implemented, including calculations of the Pearson correlation coefficient, the use of the t-distribution to test the validity of correlation between the two data sets. Also, the p value was calculated to provide the confidence level of the correlation hypothesis and its significance.
  • No Thumbnail Available
    Item
    Trust and Anti-Autonomy Modelling of Autonomous Systems
    (North Dakota State University, 2020) Rastogi, Aakanksha
    Human trust in autonomous vehicles is built upon their safe and secure operability in the most ethical, law abiding manner possible. Despite the technological advancements that autonomous vehicles are equipped with, their perplexing operation on roads often give away telltale signs of underlying vulnerabilities to threats and attack strategies which can flag their anti-autonomous traits. Anti-autonomy refers to any conduct of autonomous vehicles that goes against the principles of autonomy and subsequently resulting in their immobilized operations during unexpected roadway situations. The concept of trust is fluid, which is made complicated by anti-autonomous behavior of autonomous vehicles and affects the dimensions of intentionality, human interaction, and adoption of autonomous vehicles. Trust is impacted by intentionality, safety and risks associated with autonomous vehicles and their overall perception by human drivers, pedestrians and bicyclist sharing the roads with them. The presence of collision data involving human drivers of other cars, pedestrian, bicyclists, resulting in injuries and damages poses a significant negative impact on trust in autonomous vehicle technology. This dissertation presents and evaluates a new and innovative anti-autonomy NoTrust Artificial Neural Network model by utilizing collision data reports involving autonomous vehicles provided by California DMV from October 2014 to March 2020, which is the latest reported data. This data was augmented, labelled, classified, pre-processed, and then applied towards creation of the NoTrust ANN model using linear sequential model libraries in Keras over Tensorflow. This model was used to predict trust in autonomous vehicles. The trained model was able to achieve 100% accuracy, which was evident in the results of model compilation and training, plots of validation and training accuracies and losses. Model evaluations and predictions were used to comprehend characteristics of trust, intentionality and anti-autonomy and helped establish a relationship between them and reflected inter-dependencies among trust, intentionality, anti-autonomy, risk, and safety. Additional analyses of collision reports data was performed and the impact of several contributing factors of collisions such as vehicle driving mode, damages sustained by the vehicle, pedestrian and bicyclist involved in collisions, weather conditions, roadway surface, lighting conditions, movement of vehicle preceding collision and type of collisions was illustrated.