摘要: |
The purpose of this Innovations Deserving Exploratory Analysis (IDEA) project was to develop technology for robust traffic monitoring and automated vehicle control using decision theory and probability. It is shown that high-level traffic monitoring situations can be modeled using modern techniques, and that solving such models in real time is computationally feasible. Specifically, the focus is on models supporting the task of "plan recognition in highway environments". These models are used to infer the intended behavior of vehicles in traffic based on movement patterns and highway tactics (e.g., lane changes). The models have been developed and encoded using off-the-shelf software, in particular the Hugin (Trademark) system. Using existing algorithms and some enhancements developed as part of this project, all of the models can be solved without prohibitive computational resources. For example, real-time performance is easily achievable using hardware based on Pentium-class processors. This capability is demonstrated using Hugin (Trademark) in a standalone setup as well as in a driving simulator integrated with the SmartPATH animation system for real-time visualization of traffic scenarios. |