Optimal 3D selection technique assignment using real-time contextual analysis

Selection in 3D virtual environments can vary wildly depending on the context of the selection. Various scene attributes such as object velocity and scene density will likely impact the user's ability to accurately select an object. While there are many existing 3D selection techniques that have been well studied, they all tend to be tailored to work best in a particular set of conditions, and may not perform well when these conditions are not met. As a result, designers must compromise by taking a holistic approach to choosing a primary technique; one which works well overall, but is possibly lacking in at least one scenario. We present a software framework that allows a flexible method of leveraging several selection techniques, each performing well under certain conditions. From these, the best one is utilized at any given moment to provide the user with an optimal selection experience across more scenarios and conditions. We performed a user study comparing our framework to two common 3D selection techniques, Bendcast and Expand. We evaluated the techniques across three levels of scene density and three levels of object velocity, collecting accuracy and timing data across a large sample of participants. From our results, we were able to conclude that our auto-selection technique approach is promising but there are several characteristics of the auto-selection process that can introduce drawbacks which need to be addressed and minimized.