Advancing Transportation Research via Sensing, Control, and Visualization

Project summary:

Interpreting multidimensional data from multiple sensors (trajectories, accelerations, near misses, etc.) and evaluating the impact of various control patterns on traffic management will require a new ability to see, query, and compare the resulting space-time datasets. In other contexts (e.g., visualizing human biomechanics), the researchers have developed interactive computer graphics algorithms for displaying motion trajectories that not only convey the path of movement, but also utilize color, texture, 3D form, and animation to elucidate relationships among additional data variables (e.g. time, orientation, acceleration, proximity).

The current research project extends this visualization framework to support both data exposition and interactive data exploration. Leveraging the researchers' recently developed method for unsupervised clustering of vehicle trajectories with widely varying lengths, the new visualizations will reveal hierarchical multivariate relationships in the data (e.g., at both the individual vehicle level and at the cluster level). Side-by-side and overlay visualizations will be used to highlight changes in patterns over time, including analysis of change before and after implementing a new control module and comparison of experimental and theoretical results. Underlying automated data analysis algorithms (e.g., analysis of variance and uncertainty within trajectory clusters, analysis of significant change) will be used to target the visualizations toward the most important and useful data attributes to convey to the user. Extending recent paradigms in visual analytics, the data visualizations will be useful not just as static displays, but also as interactive data querying tools that make it possible to both test hypotheses and generate new hypotheses through data exploration. New interfaces will support interactive selection and data querying for vehicle trajectories and trajectory clusters as well as interactive navigation through large space-time datasets.