Information Framework for Transportation Applications, 10-R8175
Steven W. Dellenback, Ph.D., PMP
Inclusive Dates: 07/01/10 – 06/30/11
Background — In the transportation domain there are a large number of sources from which decisions are made. Consider the following examples:
A traditional manned vehicle 20 years ago: a driver had to rely on his vision and wits to determine the safest and most direct path to get between two points.
A traditional manned vehicle today: a driver has navigation systems, radio feeds, real-time traffic conditions on their PDA and/or Internet connection as well as potentially other sensors in their vehicle to assist in the safest and quickest path between two points.
An unmanned vehicle of several years ago: uses sensors on the vehicle to determine the "state" of the world around the vehicle and navigates between two points utilizing pre-programmed "way points."
An unmanned vehicle of today: utilizes data from other vehicles and/or the roadside to more accurately determine the optimal path.
While the Global Positioning System is important to locating a vehicle, there is a major push by the military to have their intelligent vehicles (whether manned or not) operate in "GPS-denied" environments, that is, use their surroundings to determine their location and not rely on specific GPS data. This project developed the algorithms and techniques to create a world-view model combining a number of data sources, both static and dynamic, to create a world model. Because of the huge amounts of data, an important component of the research was developing an algorithm that extracts regions of interest that are limited to what data is required by the application requesting data.
Approach — Sources of data (i.e., the proposed research is not creating this data) that are placed into the model include:
Data generated from processing the data from sensors on vehicles.
Traffic condition data available as data from Internet sources, radio broadcasts (available on a data feed in parallel with voice data).
Traffic management center data.
Landmarks such as buildings, stadiums, etc.
The internal structure of the model is based on a coordinate system, but the data can be extracted either in spatial (feature) or geolocation form.
Accomplishments — The program has successfully demonstrated the extraction of features from vision and LIDAR data to determine relative position of the vehicle.